By task format and kind of work

How entries score on the Geospatial Agent Index's tasks, grouped by task format, kind of geospatial work, data type and task length. These scores do not enter the index.

Ongoing evaluation: results update as runs complete. Last updated 16 October 2026.

Skip to results

Models68 of 68 models

Scores

By task format

Multi-step analysis score

Share of 79 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Multi-step analysis score

Multi-step analysis score
Model	Developer	Multi-step analysis score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	70	Complete	79 of 79 tasks (37 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	70	Complete	79 of 79 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	70	Complete	79 of 79 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	76	Complete	79 of 79 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	70	Complete	79 of 79 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	43	Complete	79 of 79 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	32	Complete	79 of 79 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	57	Still to run: 2 of 79 tasks	77 of 79 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	58	Complete	79 of 79 tasks (7 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	58	Complete	79 of 79 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	29	Complete	79 of 79 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	58	Complete	79 of 79 tasks (19 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	61	Complete	79 of 79 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	41	Complete	79 of 79 tasks
GLM-5.2 (high), Terminus-2	Z.ai	77	Complete	79 of 79 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	70	Complete	79 of 79 tasks
GLM-5.2 (max), Terminus-2	Z.ai	81	Complete	79 of 79 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	65	Complete	79 of 79 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	63	Complete	79 of 79 tasks
GLM-5.3 (low), Terminus-2	Z.ai	68	Complete	79 of 79 tasks
GLM-5.3 (max), Terminus-2	Z.ai	72	Complete	79 of 79 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	18	Complete	79 of 79 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	34	Complete	79 of 79 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	16	Complete	79 of 79 tasks (1 attempt not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	68	Complete	79 of 79 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	67	Complete	79 of 79 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	75	Complete	79 of 79 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	1	Complete	79 of 79 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 79 of 79 tasks	0 of 79 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	49	Complete	79 of 79 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	41	Complete	79 of 79 tasks
GLM-5.2 (max), Claude Code	Z.ai	78	Complete	79 of 79 tasks (2 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	77	Still to run: 22 of 79 tasks	57 of 79 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	70	Complete	79 of 79 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	59	Still to run: 1 of 79 tasks	78 of 79 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	73	Still to run: 35 of 79 tasks	44 of 79 tasks (35 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	49	Complete	79 of 79 tasks
gpt-oss-20b (high), Codex	OpenAI	4	Complete	79 of 79 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	70	Complete	79 of 79 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	75	Complete	79 of 79 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	78	Complete	79 of 79 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	78	Complete	79 of 79 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	76	Complete	79 of 79 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	82	Complete	79 of 79 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	68	Complete	79 of 79 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	72	Complete	79 of 79 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	84	Complete	79 of 79 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	82	Complete	79 of 79 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	76	Complete	79 of 79 tasks
Claude Opus 5 (high), Claude Code	Anthropic	75	Complete	79 of 79 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	84	Complete	79 of 79 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	73	Complete	79 of 79 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	75	Complete	79 of 79 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	72	Complete	79 of 79 tasks
GPT-6.1 Sol (high), Codex	OpenAI	71	Complete	79 of 79 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	76	Complete	79 of 79 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	70	Complete	79 of 79 tasks
GPT-6 Sol (medium), Codex	OpenAI	80	Still to run: 74 of 79 tasks	5 of 79 tasks
GPT-6 Sol (high), Codex	OpenAI	75	Complete	79 of 79 tasks
GPT-6 Luna (medium), Codex	OpenAI	69	Still to run: 20 of 79 tasks	59 of 79 tasks
GPT-6 Luna (high), Codex	OpenAI	75	Complete	79 of 79 tasks
GPT-5.6 Sol (high), Codex	OpenAI	80	Complete	79 of 79 tasks
GPT-5.6 Terra (high), Codex	OpenAI	71	Still to run: 1 of 79 tasks	78 of 79 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	71	Still to run: 58 of 79 tasks	21 of 79 tasks
GPT-5.6 Luna (high), Codex	OpenAI	82	Still to run: 14 of 79 tasks	65 of 79 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	70	Still to run: 69 of 79 tasks	10 of 79 tasks
GPT-6 Luna (low), Codex	OpenAI	69	Still to run: 25 of 79 tasks	54 of 79 tasks
GPT-5.6 Luna (low), Codex	OpenAI	69	Still to run: 50 of 79 tasks	29 of 79 tasks

Single-answer questions score

Share of 349 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Single-answer questions score

Single-answer questions score
Model	Developer	Single-answer questions score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	58	Still to run: 311 of 349 tasks	38 of 349 tasks (16 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	75	Complete	349 of 349 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	77	Complete	349 of 349 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	75	Complete	349 of 349 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	74	Complete	349 of 349 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	45	Still to run: 311 of 349 tasks	38 of 349 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	34	Still to run: 311 of 349 tasks	38 of 349 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	61	Still to run: 311 of 349 tasks	38 of 349 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	53	Still to run: 311 of 349 tasks	38 of 349 tasks (1 attempt not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	58	Still to run: 311 of 349 tasks	38 of 349 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	42	Still to run: 311 of 349 tasks	38 of 349 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	55	Still to run: 311 of 349 tasks	38 of 349 tasks (12 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	56	Complete	349 of 349 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	44	Complete	349 of 349 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (high), Terminus-2	Z.ai	75	Complete	349 of 349 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	68	Complete	349 of 349 tasks
GLM-5.2 (max), Terminus-2	Z.ai	77	Complete	349 of 349 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	68	Complete	349 of 349 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Terminus-2	Z.ai	70	Complete	349 of 349 tasks
GLM-5.3 (low), Terminus-2	Z.ai	72	Complete	349 of 349 tasks
GLM-5.3 (max), Terminus-2	Z.ai	76	Complete	349 of 349 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	29	Complete	349 of 349 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	45	Complete	349 of 349 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	31	Complete	349 of 349 tasks (5 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	74	Complete	349 of 349 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	78	Complete	349 of 349 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	78	Complete	349 of 349 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	9	Complete	349 of 349 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	37	Still to run: 322 of 349 tasks	27 of 349 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	63	Complete	349 of 349 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	44	Complete	349 of 349 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (max), Claude Code	Z.ai	79	Still to run: 2 of 349 tasks	347 of 349 tasks (5 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	74	Still to run: 9 of 349 tasks	340 of 349 tasks (6 attempts not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	79	Complete	349 of 349 tasks (1 attempt not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	68	Still to run: 1 of 349 tasks	348 of 349 tasks (68 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	51	Still to run: 314 of 349 tasks	35 of 349 tasks (52 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	35	Complete	349 of 349 tasks (7 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	10	Complete	349 of 349 tasks (20 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	74	Complete	349 of 349 tasks (14 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	78	Complete	349 of 349 tasks (3 attempts not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	84	Complete	349 of 349 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	87	Complete	349 of 349 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	72	Complete	349 of 349 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	79	Complete	349 of 349 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	76	Complete	349 of 349 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	80	Complete	349 of 349 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	88	Complete	349 of 349 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	81	Complete	349 of 349 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	77	Complete	349 of 349 tasks
Claude Opus 5 (high), Claude Code	Anthropic	85	Complete	349 of 349 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	74	Complete	349 of 349 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	80	Complete	349 of 349 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	76	Complete	349 of 349 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	86	Complete	349 of 349 tasks
GPT-6.1 Sol (high), Codex	OpenAI	85	Complete	349 of 349 tasks (1 attempt not counted: provider or infrastructure error)
GPT-6.1 Sol (xhigh), Codex	OpenAI	86	Complete	349 of 349 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	74	Complete	349 of 349 tasks
GPT-6 Sol (medium), Codex	OpenAI	86	Still to run: 165 of 349 tasks	184 of 349 tasks
GPT-6 Sol (high), Codex	OpenAI	84	Complete	349 of 349 tasks
GPT-6 Luna (medium), Codex	OpenAI	61	Complete	349 of 349 tasks
GPT-6 Luna (high), Codex	OpenAI	73	Complete	349 of 349 tasks
GPT-5.6 Sol (high), Codex	OpenAI	81	Complete	349 of 349 tasks
GPT-5.6 Terra (high), Codex	OpenAI	74	Complete	349 of 349 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	73	Still to run: 48 of 349 tasks	301 of 349 tasks
GPT-5.6 Luna (high), Codex	OpenAI	76	Complete	349 of 349 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	84	Still to run: 157 of 349 tasks	192 of 349 tasks
GPT-6 Luna (low), Codex	OpenAI	54	Still to run: 1 of 349 tasks	348 of 349 tasks
GPT-5.6 Luna (low), Codex	OpenAI	66	Still to run: 5 of 349 tasks	344 of 349 tasks

Recognising infeasible requests score

Share of 60 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Recognising infeasible requests score

Recognising infeasible requests score
Model	Developer	Recognising infeasible requests score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	40	Complete	60 of 60 tasks (26 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	53	Complete	60 of 60 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	58	Complete	60 of 60 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	55	Complete	60 of 60 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	62	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	62	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	65	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	62	Complete	60 of 60 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	93	Complete	60 of 60 tasks (6 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	97	Complete	60 of 60 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	82	Complete	60 of 60 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	100	Complete	60 of 60 tasks (13 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63	Complete	60 of 60 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43	Complete	60 of 60 tasks
GLM-5.2 (high), Terminus-2	Z.ai	53	Complete	60 of 60 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	45	Complete	60 of 60 tasks
GLM-5.2 (max), Terminus-2	Z.ai	48	Complete	60 of 60 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	47	Complete	60 of 60 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	48	Complete	60 of 60 tasks
GLM-5.3 (low), Terminus-2	Z.ai	53	Complete	60 of 60 tasks
GLM-5.3 (max), Terminus-2	Z.ai	55	Complete	60 of 60 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	60	Complete	60 of 60 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	58	Complete	60 of 60 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	67	Complete	60 of 60 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	57	Complete	60 of 60 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	57	Complete	60 of 60 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	50	Complete	60 of 60 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	13	Complete	60 of 60 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 60 of 60 tasks	0 of 60 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	53	Complete	60 of 60 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	40	Complete	60 of 60 tasks
GLM-5.2 (max), Claude Code	Z.ai	52	Complete	60 of 60 tasks (4 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	65	Still to run: 8 of 60 tasks	52 of 60 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	58	Complete	60 of 60 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	32	Complete	60 of 60 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	65	Still to run: 23 of 60 tasks	37 of 60 tasks (22 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	62	Complete	60 of 60 tasks
gpt-oss-20b (high), Codex	OpenAI	17	Complete	60 of 60 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	58	Complete	60 of 60 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	50	Complete	60 of 60 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	80	Complete	60 of 60 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	78	Complete	60 of 60 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	75	Complete	60 of 60 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	80	Complete	60 of 60 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	82	Complete	60 of 60 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	83	Complete	60 of 60 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	83	Complete	60 of 60 tasks
Claude Opus 5 (high), Claude Code	Anthropic	78	Complete	60 of 60 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	72	Complete	60 of 60 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	70	Complete	60 of 60 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	90	Complete	60 of 60 tasks
GPT-6.1 Sol (high), Codex	OpenAI	87	Complete	60 of 60 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	87	Complete	60 of 60 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	80	Complete	60 of 60 tasks
GPT-6 Sol (medium), Codex	OpenAI	50	Still to run: 56 of 60 tasks	4 of 60 tasks
GPT-6 Sol (high), Codex	OpenAI	78	Complete	60 of 60 tasks
GPT-6 Luna (medium), Codex	OpenAI	38	Still to run: 8 of 60 tasks	52 of 60 tasks
GPT-6 Luna (high), Codex	OpenAI	75	Complete	60 of 60 tasks
GPT-5.6 Sol (high), Codex	OpenAI	58	Complete	60 of 60 tasks
GPT-5.6 Terra (high), Codex	OpenAI	60	Complete	60 of 60 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	41	Still to run: 43 of 60 tasks	17 of 60 tasks
GPT-5.6 Luna (high), Codex	OpenAI	53	Still to run: 5 of 60 tasks	55 of 60 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	50	Still to run: 52 of 60 tasks	8 of 60 tasks
GPT-6 Luna (low), Codex	OpenAI	30	Still to run: 13 of 60 tasks	47 of 60 tasks
GPT-5.6 Luna (low), Codex	OpenAI	50	Still to run: 36 of 60 tasks	24 of 60 tasks

Score by task format (0 to 100) and the Geospatial Agent Index
Model	Multi-step analysis (79 tasks)	Single-answer questions (349 tasks)	Recognising infeasible requests (60 tasks)	Geospatial Agent Index
Claude Opus 5.5 (xhigh), Claude Code	84	88	82	85
Claude Opus 5.5 (high), Claude Code	78	87	78	82
GPT-6.1 Sol (xhigh), Codex	76	86	87	82
GPT-6.1 Sol (medium), Codex	72	86	90	82
Claude Opus 5.5 (medium), Claude Code	78	84	80	80
GPT-6.1 Sol (high), Codex	71	85	87	80
Claude Opus 5 (high), Claude Code	75	85	78	80
GPT-6 Sol (high), Codex	75	84	78	80
Claude Sonnet 5.5 (xhigh), Claude Code	82	81	83	79
Claude Sonnet 5.5 (high), Claude Code	82	79	80	78
Claude Opus 4.8 (high), Claude Code	73	80	72	76
Claude Haiku 5.5 (xhigh), Claude Code	76	77	83	76
Claude Haiku 5.5 (high), Claude Code	72	80	73	76
GPT-5.6 Sol (high), Codex	80	81	58	75
Claude Sonnet 5 (high), Claude Code	84	74	73	74
Claude Opus 4.7 (high), Claude Code	75	76	70	74
GLM-5.3 (max), Claude Code	70	79	58	73
GPT-6 Luna (xhigh), Codex	70	74	80	73
Claude Haiku 5.5 (medium), Claude Code	68	76	73	72
Kimi K3 (Reasoning), Terminus-2	75	78	50	72
DeepSeek V4 Flash (high), Terminus-2	70	77	58	72
Kimi K2.7 Code (Reasoning), Kimi Code	75	78	50	72
GLM-5.2 (max), Terminus-2	81	77	48	72
GPT-6 Luna (high), Codex	75	73	75	71
Kimi K2.7 Code (Reasoning), Terminus-2	67	78	57	71
DeepSeek V4 Pro (max), Terminus-2	76	75	55	71
Claude Sonnet 5.5 (medium), Claude Code	76	72	75	71
GLM-5.2 (high), Terminus-2	77	75	53	71
GLM-5.3 (max), Terminus-2	72	76	55	70
DeepSeek V4 Pro (high), Terminus-2	70	74	62	70
Kimi K2.6 (Reasoning), Terminus-2	68	74	57	70
DeepSeek V4 Flash (max), Terminus-2	70	75	53	69
Kimi K2.6 (Reasoning), Kimi Code	70	74	58	69
GLM-5.3 (low), Terminus-2	68	72	53	66
GLM-5.2 (Non-reasoning), Terminus-2	70	68	45	63
GLM-5.3-Flash (max), Terminus-2	63	70	48	63
GLM-5.3-Flash (low), Terminus-2	65	68	47	62
Claude Haiku 4.5 (Reasoning), Claude Code	49	63	53	58
Gemma 4 26B A4B (Reasoning), Terminus-2	61	56	63	56
gpt-oss-120b (high), Terminus-2	34	45	58	46
GLM-4.7-Flash (Reasoning), Terminus-2	41	44	43	43
gpt-oss-120b (high), Codex	49	35	62	43
GLM-4.7-Flash (Reasoning), Claude Code	41	44	40	41
gpt-oss-20b (high), Terminus-2	16	31	67	33
gpt-oss-120b (low), Terminus-2	18	29	60	31
gpt-oss-20b (high), Codex	4	10	17	9
Llama 3.3 70B (Non-reasoning), Terminus-2	1	9	13	8
GPT-6 Sol (medium), Codex	80	86	50	77, not every task attempted yet (193 of 488 tasks attempted)
GPT-5.6 Luna (xhigh), Codex	70	84	50	77, not every task attempted yet (210 of 488 tasks attempted)
GLM-5.2 (max), Claude Code	78	79	52	73, not every task attempted yet
GPT-5.6 Luna (high), Codex	82	76	53	72, not every task attempted yet (469 of 488 tasks attempted)
GLM-5.3-Flash (max), Claude Code	77	74	65	72, not every task attempted yet (449 of 488 tasks attempted)
GPT-5.6 Terra (high), Codex	71	74	60	69, not every task attempted yet
Gemini 3.8 Flash (low), Terminus-2	58	55	100	67, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (high), Terminus-2	58	58	97	66, not every task attempted yet (177 of 488 tasks attempted)
GPT-5.6 Luna (medium), Codex	71	73	41	66, not every task attempted yet (339 of 488 tasks attempted)
DeepSeek V4 Pro (high), Codex	73	51	65	62, not every task attempted yet (116 of 488 tasks attempted)
GPT-5.6 Luna (low), Codex	69	66	50	60, not every task attempted yet (397 of 488 tasks attempted)
DeepSeek V4 Flash (high), Codex	59	68	32	59, not every task attempted yet
Claude Sonnet 4.6 (high), Terminus-2	70	58	40	58, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (low), Terminus-2	58	53	93	57, not every task attempted yet (177 of 488 tasks attempted)
GPT-6 Luna (medium), Codex	69	61	38	57, not every task attempted yet (460 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (high), Terminus-2	57	61	62	56, not every task attempted yet (175 of 488 tasks attempted)
GPT-6 Luna (low), Codex	69	54	30	51, not every task attempted yet (449 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (low), Terminus-2	43	45	62	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.8 Flash (high), Terminus-2	29	42	82	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (minimal), Terminus-2	32	34	65	42, not every task attempted yet (177 of 488 tasks attempted)
Qwen 3.8 27B (xhigh), Terminus-2	No data	37	No data	37, not every task attempted yet (27 of 488 tasks attempted)

* Not every task attempted yet: the score may still change. What the asterisk means.

By kind of work

Raster and remote sensing score

Share of 181 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Raster and remote sensing score

Raster and remote sensing score
Model	Developer	Raster and remote sensing score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	58	Still to run: 150 of 181 tasks	31 of 181 tasks (14 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	78	Complete	181 of 181 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	78	Complete	181 of 181 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	76	Complete	181 of 181 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	77	Complete	181 of 181 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	39	Still to run: 150 of 181 tasks	31 of 181 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	29	Still to run: 150 of 181 tasks	31 of 181 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	52	Still to run: 150 of 181 tasks	31 of 181 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	55	Still to run: 150 of 181 tasks	31 of 181 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	61	Still to run: 150 of 181 tasks	31 of 181 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	35	Still to run: 150 of 181 tasks	31 of 181 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	61	Still to run: 150 of 181 tasks	31 of 181 tasks (8 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	59	Complete	181 of 181 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42	Complete	181 of 181 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (high), Terminus-2	Z.ai	77	Complete	181 of 181 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	72	Complete	181 of 181 tasks
GLM-5.2 (max), Terminus-2	Z.ai	76	Complete	181 of 181 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	67	Complete	181 of 181 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	67	Complete	181 of 181 tasks
GLM-5.3 (low), Terminus-2	Z.ai	74	Complete	181 of 181 tasks
GLM-5.3 (max), Terminus-2	Z.ai	76	Complete	181 of 181 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	25	Complete	181 of 181 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	44	Complete	181 of 181 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	33	Complete	181 of 181 tasks (4 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	77	Complete	181 of 181 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	76	Complete	181 of 181 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	80	Complete	181 of 181 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	7	Complete	181 of 181 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	38	Still to run: 165 of 181 tasks	16 of 181 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	65	Complete	181 of 181 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	41	Complete	181 of 181 tasks
GLM-5.2 (max), Claude Code	Z.ai	80	Still to run: 1 of 181 tasks	180 of 181 tasks (5 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	74	Still to run: 11 of 181 tasks	170 of 181 tasks
GLM-5.3 (max), Claude Code	Z.ai	80	Complete	181 of 181 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	68	Still to run: 1 of 181 tasks	180 of 181 tasks (36 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	62	Still to run: 155 of 181 tasks	26 of 181 tasks (26 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	40	Complete	181 of 181 tasks (5 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	11	Complete	181 of 181 tasks (11 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	75	Complete	181 of 181 tasks (5 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	80	Complete	181 of 181 tasks (2 attempts not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	83	Complete	181 of 181 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	88	Complete	181 of 181 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	71	Complete	181 of 181 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	80	Complete	181 of 181 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	76	Complete	181 of 181 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	79	Complete	181 of 181 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	89	Complete	181 of 181 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	81	Complete	181 of 181 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	76	Complete	181 of 181 tasks
Claude Opus 5 (high), Claude Code	Anthropic	83	Complete	181 of 181 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	76	Complete	181 of 181 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	82	Complete	181 of 181 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	79	Complete	181 of 181 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	83	Complete	181 of 181 tasks
GPT-6.1 Sol (high), Codex	OpenAI	83	Complete	181 of 181 tasks (1 attempt not counted: provider or infrastructure error)
GPT-6.1 Sol (xhigh), Codex	OpenAI	85	Complete	181 of 181 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	73	Complete	181 of 181 tasks
GPT-6 Sol (medium), Codex	OpenAI	87	Still to run: 102 of 181 tasks	79 of 181 tasks
GPT-6 Sol (high), Codex	OpenAI	83	Complete	181 of 181 tasks
GPT-6 Luna (medium), Codex	OpenAI	63	Still to run: 1 of 181 tasks	180 of 181 tasks
GPT-6 Luna (high), Codex	OpenAI	73	Complete	181 of 181 tasks
GPT-5.6 Sol (high), Codex	OpenAI	78	Complete	181 of 181 tasks
GPT-5.6 Terra (high), Codex	OpenAI	76	Complete	181 of 181 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	78	Still to run: 43 of 181 tasks	138 of 181 tasks
GPT-5.6 Luna (high), Codex	OpenAI	78	Still to run: 1 of 181 tasks	180 of 181 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	82	Still to run: 104 of 181 tasks	77 of 181 tasks
GPT-6 Luna (low), Codex	OpenAI	55	Still to run: 2 of 181 tasks	179 of 181 tasks
GPT-5.6 Luna (low), Codex	OpenAI	70	Still to run: 7 of 181 tasks	174 of 181 tasks

Vector and overlay score

Share of 118 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Vector and overlay score

Vector and overlay score
Model	Developer	Vector and overlay score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	40	Still to run: 113 of 118 tasks	5 of 118 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	75	Complete	118 of 118 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	77	Complete	118 of 118 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	78	Complete	118 of 118 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	73	Complete	118 of 118 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	40	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	40	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	60	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	80	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	40	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	20	Still to run: 113 of 118 tasks	5 of 118 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	0	Still to run: 113 of 118 tasks	5 of 118 tasks (1 attempt not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	57	Complete	118 of 118 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	49	Complete	118 of 118 tasks
GLM-5.2 (high), Terminus-2	Z.ai	79	Complete	118 of 118 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	65	Complete	118 of 118 tasks
GLM-5.2 (max), Terminus-2	Z.ai	82	Complete	118 of 118 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	75	Complete	118 of 118 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Terminus-2	Z.ai	86	Complete	118 of 118 tasks
GLM-5.3 (low), Terminus-2	Z.ai	75	Complete	118 of 118 tasks
GLM-5.3 (max), Terminus-2	Z.ai	80	Complete	118 of 118 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	34	Complete	118 of 118 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	41	Complete	118 of 118 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	26	Complete	118 of 118 tasks (1 attempt not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	74	Complete	118 of 118 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	81	Complete	118 of 118 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	76	Complete	118 of 118 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	12	Complete	118 of 118 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 118 of 118 tasks	0 of 118 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	67	Complete	118 of 118 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	49	Complete	118 of 118 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (max), Claude Code	Z.ai	81	Complete	118 of 118 tasks
GLM-5.3-Flash (max), Claude Code	Z.ai	79	Complete	118 of 118 tasks
GLM-5.3 (max), Claude Code	Z.ai	81	Complete	118 of 118 tasks (1 attempt not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	75	Complete	118 of 118 tasks (6 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	0	Still to run: 116 of 118 tasks	2 of 118 tasks (3 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	23	Complete	118 of 118 tasks (2 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	12	Complete	118 of 118 tasks (9 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	77	Complete	118 of 118 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	77	Complete	118 of 118 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	92	Complete	118 of 118 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	89	Complete	118 of 118 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	81	Complete	118 of 118 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	86	Complete	118 of 118 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	80	Complete	118 of 118 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	82	Complete	118 of 118 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	94	Complete	118 of 118 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	89	Complete	118 of 118 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	82	Complete	118 of 118 tasks
Claude Opus 5 (high), Claude Code	Anthropic	88	Complete	118 of 118 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	75	Complete	118 of 118 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	80	Complete	118 of 118 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	75	Complete	118 of 118 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	94	Complete	118 of 118 tasks
GPT-6.1 Sol (high), Codex	OpenAI	92	Complete	118 of 118 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	93	Complete	118 of 118 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	78	Complete	118 of 118 tasks
GPT-6 Sol (medium), Codex	OpenAI	86	Still to run: 35 of 118 tasks	83 of 118 tasks
GPT-6 Sol (high), Codex	OpenAI	86	Complete	118 of 118 tasks
GPT-6 Luna (medium), Codex	OpenAI	65	Complete	118 of 118 tasks
GPT-6 Luna (high), Codex	OpenAI	77	Complete	118 of 118 tasks
GPT-5.6 Sol (high), Codex	OpenAI	87	Complete	118 of 118 tasks
GPT-5.6 Terra (high), Codex	OpenAI	77	Complete	118 of 118 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	76	Still to run: 3 of 118 tasks	115 of 118 tasks
GPT-5.6 Luna (high), Codex	OpenAI	79	Complete	118 of 118 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	86	Still to run: 21 of 118 tasks	97 of 118 tasks
GPT-6 Luna (low), Codex	OpenAI	57	Complete	118 of 118 tasks
GPT-5.6 Luna (low), Codex	OpenAI	72	Still to run: 1 of 118 tasks	117 of 118 tasks

Climate and time series score

Share of 59 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Climate and time series score

Climate and time series score
Model	Developer	Climate and time series score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	73	Still to run: 48 of 59 tasks	11 of 59 tasks (5 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	66	Complete	59 of 59 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	73	Complete	59 of 59 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	66	Complete	59 of 59 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	66	Complete	59 of 59 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	45	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	36	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	55	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	45	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	73	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	45	Still to run: 48 of 59 tasks	11 of 59 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	64	Still to run: 48 of 59 tasks	11 of 59 tasks (4 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	47	Complete	59 of 59 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	41	Complete	59 of 59 tasks
GLM-5.2 (high), Terminus-2	Z.ai	68	Complete	59 of 59 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	58	Complete	59 of 59 tasks
GLM-5.2 (max), Terminus-2	Z.ai	71	Complete	59 of 59 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	59	Complete	59 of 59 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	51	Complete	59 of 59 tasks
GLM-5.3 (low), Terminus-2	Z.ai	59	Complete	59 of 59 tasks
GLM-5.3 (max), Terminus-2	Z.ai	71	Complete	59 of 59 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	29	Complete	59 of 59 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	53	Complete	59 of 59 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	31	Complete	59 of 59 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	69	Complete	59 of 59 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	75	Complete	59 of 59 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	71	Complete	59 of 59 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	10	Complete	59 of 59 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	36	Still to run: 48 of 59 tasks	11 of 59 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	49	Complete	59 of 59 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	39	Complete	59 of 59 tasks
GLM-5.2 (max), Claude Code	Z.ai	72	Still to run: 1 of 59 tasks	58 of 59 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	67	Still to run: 2 of 59 tasks	57 of 59 tasks (6 attempts not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	76	Complete	59 of 59 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	58	Complete	59 of 59 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	42	Still to run: 47 of 59 tasks	12 of 59 tasks (27 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	44	Complete	59 of 59 tasks
gpt-oss-20b (high), Codex	OpenAI	5	Complete	59 of 59 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	63	Complete	59 of 59 tasks (9 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	73	Complete	59 of 59 tasks (1 attempt not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	73	Complete	59 of 59 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	75	Complete	59 of 59 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	61	Complete	59 of 59 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	69	Complete	59 of 59 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	66	Complete	59 of 59 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	76	Complete	59 of 59 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	75	Complete	59 of 59 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	66	Complete	59 of 59 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	69	Complete	59 of 59 tasks
Claude Opus 5 (high), Claude Code	Anthropic	78	Complete	59 of 59 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	69	Complete	59 of 59 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	73	Complete	59 of 59 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	71	Complete	59 of 59 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	75	Complete	59 of 59 tasks
GPT-6.1 Sol (high), Codex	OpenAI	75	Complete	59 of 59 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	75	Complete	59 of 59 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	73	Complete	59 of 59 tasks
GPT-6 Sol (medium), Codex	OpenAI	83	Still to run: 35 of 59 tasks	24 of 59 tasks
GPT-6 Sol (high), Codex	OpenAI	80	Complete	59 of 59 tasks
GPT-6 Luna (medium), Codex	OpenAI	47	Complete	59 of 59 tasks
GPT-6 Luna (high), Codex	OpenAI	66	Complete	59 of 59 tasks
GPT-5.6 Sol (high), Codex	OpenAI	76	Complete	59 of 59 tasks
GPT-5.6 Terra (high), Codex	OpenAI	61	Complete	59 of 59 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	58	Still to run: 6 of 59 tasks	53 of 59 tasks
GPT-5.6 Luna (high), Codex	OpenAI	69	Complete	59 of 59 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	90	Still to run: 38 of 59 tasks	21 of 59 tasks
GPT-6 Luna (low), Codex	OpenAI	49	Complete	59 of 59 tasks
GPT-5.6 Luna (low), Codex	OpenAI	42	Complete	59 of 59 tasks

Spatial statistics and interpolation score

Share of 13 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Spatial statistics and interpolation score

Spatial statistics and interpolation score
Model	Developer	Spatial statistics and interpolation score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	46	Complete	13 of 13 tasks (7 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	23	Complete	13 of 13 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	38	Complete	13 of 13 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	46	Complete	13 of 13 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	23	Complete	13 of 13 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	8	Complete	13 of 13 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	0	Complete	13 of 13 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	31	Complete	13 of 13 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	46	Complete	13 of 13 tasks (1 attempt not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	54	Complete	13 of 13 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	31	Complete	13 of 13 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	62	Complete	13 of 13 tasks (3 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	31	Complete	13 of 13 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	15	Complete	13 of 13 tasks
GLM-5.2 (high), Terminus-2	Z.ai	46	Complete	13 of 13 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	31	Complete	13 of 13 tasks
GLM-5.2 (max), Terminus-2	Z.ai	62	Complete	13 of 13 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	38	Complete	13 of 13 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	38	Complete	13 of 13 tasks
GLM-5.3 (low), Terminus-2	Z.ai	23	Complete	13 of 13 tasks
GLM-5.3 (max), Terminus-2	Z.ai	54	Complete	13 of 13 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	8	Complete	13 of 13 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	15	Complete	13 of 13 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	0	Complete	13 of 13 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	15	Complete	13 of 13 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	38	Complete	13 of 13 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	54	Complete	13 of 13 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	0	Complete	13 of 13 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 13 of 13 tasks	0 of 13 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	23	Complete	13 of 13 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	23	Complete	13 of 13 tasks
GLM-5.2 (max), Claude Code	Z.ai	46	Complete	13 of 13 tasks
GLM-5.3-Flash (max), Claude Code	Z.ai	78	Still to run: 4 of 13 tasks	9 of 13 tasks
GLM-5.3 (max), Claude Code	Z.ai	38	Complete	13 of 13 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	33	Still to run: 1 of 13 tasks	12 of 13 tasks (6 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	17	Still to run: 7 of 13 tasks	6 of 13 tasks (7 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	0	Complete	13 of 13 tasks
gpt-oss-20b (high), Codex	OpenAI	0	Complete	13 of 13 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	54	Complete	13 of 13 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	31	Complete	13 of 13 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	46	Complete	13 of 13 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	54	Complete	13 of 13 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	62	Complete	13 of 13 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	62	Complete	13 of 13 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	23	Complete	13 of 13 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	31	Complete	13 of 13 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	62	Complete	13 of 13 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	46	Complete	13 of 13 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	46	Complete	13 of 13 tasks
Claude Opus 5 (high), Claude Code	Anthropic	46	Complete	13 of 13 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	54	Complete	13 of 13 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	38	Complete	13 of 13 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	38	Complete	13 of 13 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	31	Complete	13 of 13 tasks
GPT-6.1 Sol (high), Codex	OpenAI	31	Complete	13 of 13 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	38	Complete	13 of 13 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	46	Complete	13 of 13 tasks
GPT-6 Sol (medium), Codex	OpenAI	50	Still to run: 11 of 13 tasks	2 of 13 tasks
GPT-6 Sol (high), Codex	OpenAI	38	Complete	13 of 13 tasks
GPT-6 Luna (medium), Codex	OpenAI	22	Still to run: 4 of 13 tasks	9 of 13 tasks
GPT-6 Luna (high), Codex	OpenAI	54	Complete	13 of 13 tasks
GPT-5.6 Sol (high), Codex	OpenAI	62	Complete	13 of 13 tasks
GPT-5.6 Terra (high), Codex	OpenAI	50	Still to run: 1 of 13 tasks	12 of 13 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	0	Still to run: 9 of 13 tasks	4 of 13 tasks
GPT-5.6 Luna (high), Codex	OpenAI	40	Still to run: 3 of 13 tasks	10 of 13 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	33	Still to run: 10 of 13 tasks	3 of 13 tasks
GPT-6 Luna (low), Codex	OpenAI	38	Still to run: 5 of 13 tasks	8 of 13 tasks
GPT-5.6 Luna (low), Codex	OpenAI	0	Still to run: 8 of 13 tasks	5 of 13 tasks

Mapping and cartography score

Share of 57 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Mapping and cartography score

Mapping and cartography score
Model	Developer	Mapping and cartography score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	75	Complete	57 of 57 tasks (25 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	79	Complete	57 of 57 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	75	Complete	57 of 57 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	82	Complete	57 of 57 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	81	Complete	57 of 57 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	54	Complete	57 of 57 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	40	Complete	57 of 57 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	69	Still to run: 2 of 57 tasks	55 of 57 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	60	Complete	57 of 57 tasks (5 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	56	Complete	57 of 57 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	32	Complete	57 of 57 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	58	Complete	57 of 57 tasks (15 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	70	Complete	57 of 57 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	47	Complete	57 of 57 tasks
GLM-5.2 (high), Terminus-2	Z.ai	81	Complete	57 of 57 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	79	Complete	57 of 57 tasks
GLM-5.2 (max), Terminus-2	Z.ai	82	Complete	57 of 57 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	67	Complete	57 of 57 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	67	Complete	57 of 57 tasks
GLM-5.3 (low), Terminus-2	Z.ai	77	Complete	57 of 57 tasks
GLM-5.3 (max), Terminus-2	Z.ai	75	Complete	57 of 57 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	21	Complete	57 of 57 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	40	Complete	57 of 57 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	21	Complete	57 of 57 tasks (1 attempt not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	77	Complete	57 of 57 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	74	Complete	57 of 57 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	81	Complete	57 of 57 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2	Complete	57 of 57 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 57 of 57 tasks	0 of 57 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	56	Complete	57 of 57 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	49	Complete	57 of 57 tasks
GLM-5.2 (max), Claude Code	Z.ai	84	Complete	57 of 57 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	74	Still to run: 14 of 57 tasks	43 of 57 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	75	Complete	57 of 57 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	61	Complete	57 of 57 tasks (20 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	85	Still to run: 24 of 57 tasks	33 of 57 tasks (24 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	61	Complete	57 of 57 tasks
gpt-oss-20b (high), Codex	OpenAI	4	Complete	57 of 57 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	74	Complete	57 of 57 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	82	Complete	57 of 57 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	86	Complete	57 of 57 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	86	Complete	57 of 57 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	79	Complete	57 of 57 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	84	Complete	57 of 57 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	79	Complete	57 of 57 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	81	Complete	57 of 57 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	88	Complete	57 of 57 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	88	Complete	57 of 57 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	86	Complete	57 of 57 tasks
Claude Opus 5 (high), Claude Code	Anthropic	82	Complete	57 of 57 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	89	Complete	57 of 57 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	82	Complete	57 of 57 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	82	Complete	57 of 57 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	81	Complete	57 of 57 tasks
GPT-6.1 Sol (high), Codex	OpenAI	81	Complete	57 of 57 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	84	Complete	57 of 57 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	72	Complete	57 of 57 tasks
GPT-6 Sol (medium), Codex	OpenAI	100	Still to run: 56 of 57 tasks	1 of 57 tasks
GPT-6 Sol (high), Codex	OpenAI	84	Complete	57 of 57 tasks
GPT-6 Luna (medium), Codex	OpenAI	76	Still to run: 15 of 57 tasks	42 of 57 tasks
GPT-6 Luna (high), Codex	OpenAI	79	Complete	57 of 57 tasks
GPT-5.6 Sol (high), Codex	OpenAI	82	Complete	57 of 57 tasks
GPT-5.6 Terra (high), Codex	OpenAI	74	Complete	57 of 57 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	83	Still to run: 45 of 57 tasks	12 of 57 tasks
GPT-5.6 Luna (high), Codex	OpenAI	87	Still to run: 10 of 57 tasks	47 of 57 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	75	Still to run: 53 of 57 tasks	4 of 57 tasks
GPT-6 Luna (low), Codex	OpenAI	71	Still to run: 19 of 57 tasks	38 of 57 tasks
GPT-5.6 Luna (low), Codex	OpenAI	83	Still to run: 39 of 57 tasks	18 of 57 tasks

Recognising infeasible requests score

Share of 60 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Recognising infeasible requests score

Recognising infeasible requests score
Model	Developer	Recognising infeasible requests score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	40	Complete	60 of 60 tasks (26 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	53	Complete	60 of 60 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	58	Complete	60 of 60 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	55	Complete	60 of 60 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	62	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	62	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	65	Complete	60 of 60 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	62	Complete	60 of 60 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	93	Complete	60 of 60 tasks (6 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	97	Complete	60 of 60 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	82	Complete	60 of 60 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	100	Complete	60 of 60 tasks (13 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63	Complete	60 of 60 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43	Complete	60 of 60 tasks
GLM-5.2 (high), Terminus-2	Z.ai	53	Complete	60 of 60 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	45	Complete	60 of 60 tasks
GLM-5.2 (max), Terminus-2	Z.ai	48	Complete	60 of 60 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	47	Complete	60 of 60 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	48	Complete	60 of 60 tasks
GLM-5.3 (low), Terminus-2	Z.ai	53	Complete	60 of 60 tasks
GLM-5.3 (max), Terminus-2	Z.ai	55	Complete	60 of 60 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	60	Complete	60 of 60 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	58	Complete	60 of 60 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	67	Complete	60 of 60 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	57	Complete	60 of 60 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	57	Complete	60 of 60 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	50	Complete	60 of 60 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	13	Complete	60 of 60 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 60 of 60 tasks	0 of 60 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	53	Complete	60 of 60 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	40	Complete	60 of 60 tasks
GLM-5.2 (max), Claude Code	Z.ai	52	Complete	60 of 60 tasks (4 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	65	Still to run: 8 of 60 tasks	52 of 60 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	58	Complete	60 of 60 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	32	Complete	60 of 60 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	65	Still to run: 23 of 60 tasks	37 of 60 tasks (22 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	62	Complete	60 of 60 tasks
gpt-oss-20b (high), Codex	OpenAI	17	Complete	60 of 60 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	58	Complete	60 of 60 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	50	Complete	60 of 60 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	80	Complete	60 of 60 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	78	Complete	60 of 60 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	75	Complete	60 of 60 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	80	Complete	60 of 60 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	82	Complete	60 of 60 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	83	Complete	60 of 60 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	83	Complete	60 of 60 tasks
Claude Opus 5 (high), Claude Code	Anthropic	78	Complete	60 of 60 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	73	Complete	60 of 60 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	72	Complete	60 of 60 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	70	Complete	60 of 60 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	90	Complete	60 of 60 tasks
GPT-6.1 Sol (high), Codex	OpenAI	87	Complete	60 of 60 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	87	Complete	60 of 60 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	80	Complete	60 of 60 tasks
GPT-6 Sol (medium), Codex	OpenAI	50	Still to run: 56 of 60 tasks	4 of 60 tasks
GPT-6 Sol (high), Codex	OpenAI	78	Complete	60 of 60 tasks
GPT-6 Luna (medium), Codex	OpenAI	38	Still to run: 8 of 60 tasks	52 of 60 tasks
GPT-6 Luna (high), Codex	OpenAI	75	Complete	60 of 60 tasks
GPT-5.6 Sol (high), Codex	OpenAI	58	Complete	60 of 60 tasks
GPT-5.6 Terra (high), Codex	OpenAI	60	Complete	60 of 60 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	41	Still to run: 43 of 60 tasks	17 of 60 tasks
GPT-5.6 Luna (high), Codex	OpenAI	53	Still to run: 5 of 60 tasks	55 of 60 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	50	Still to run: 52 of 60 tasks	8 of 60 tasks
GPT-6 Luna (low), Codex	OpenAI	30	Still to run: 13 of 60 tasks	47 of 60 tasks
GPT-5.6 Luna (low), Codex	OpenAI	50	Still to run: 36 of 60 tasks	24 of 60 tasks

Score by kind of work (0 to 100) and the Geospatial Agent Index
Model	Raster and remote sensing (181 tasks)	Vector and overlay (118 tasks)	Climate and time series (59 tasks)	Spatial statistics and interpolation (13 tasks)	Mapping and cartography (57 tasks)	Recognising infeasible requests (60 tasks)	Geospatial Agent Index
Claude Opus 5.5 (xhigh), Claude Code	89	94	75	62	88	82	85
Claude Opus 5.5 (high), Claude Code	88	89	75	54	86	78	82
GPT-6.1 Sol (xhigh), Codex	85	93	75	38	84	87	82
GPT-6.1 Sol (medium), Codex	83	94	75	31	81	90	82
Claude Opus 5.5 (medium), Claude Code	83	92	73	46	86	80	80
GPT-6.1 Sol (high), Codex	83	92	75	31	81	87	80
Claude Opus 5 (high), Claude Code	83	88	78	46	82	78	80
GPT-6 Sol (high), Codex	83	86	80	38	84	78	80
Claude Sonnet 5.5 (xhigh), Claude Code	81	89	66	46	88	83	79
Claude Sonnet 5.5 (high), Claude Code	80	86	69	62	84	80	78
Claude Opus 4.8 (high), Claude Code	82	80	73	38	82	72	76
Claude Haiku 5.5 (xhigh), Claude Code	76	82	69	46	86	83	76
Claude Haiku 5.5 (high), Claude Code	79	82	76	31	81	73	76
GPT-5.6 Sol (high), Codex	78	87	76	62	82	58	75
Claude Sonnet 5 (high), Claude Code	76	75	69	54	89	73	74
Claude Opus 4.7 (high), Claude Code	79	75	71	38	82	70	74
GLM-5.3 (max), Claude Code	80	81	76	38	75	58	73
GPT-6 Luna (xhigh), Codex	73	78	73	46	72	80	73
Claude Haiku 5.5 (medium), Claude Code	76	80	66	23	79	73	72
Kimi K3 (Reasoning), Terminus-2	80	76	71	54	81	50	72
DeepSeek V4 Flash (high), Terminus-2	78	77	73	38	75	58	72
Kimi K2.7 Code (Reasoning), Kimi Code	80	77	73	31	82	50	72
GLM-5.2 (max), Terminus-2	76	82	71	62	82	48	72
GPT-6 Luna (high), Codex	73	77	66	54	79	75	71
Kimi K2.7 Code (Reasoning), Terminus-2	76	81	75	38	74	57	71
DeepSeek V4 Pro (max), Terminus-2	76	78	66	46	82	55	71
Claude Sonnet 5.5 (medium), Claude Code	71	81	61	62	79	75	71
GLM-5.2 (high), Terminus-2	77	79	68	46	81	53	71
GLM-5.3 (max), Terminus-2	76	80	71	54	75	55	70
DeepSeek V4 Pro (high), Terminus-2	77	73	66	23	81	62	70
Kimi K2.6 (Reasoning), Terminus-2	77	74	69	15	77	57	70
DeepSeek V4 Flash (max), Terminus-2	78	75	66	23	79	53	69
Kimi K2.6 (Reasoning), Kimi Code	75	77	63	54	74	58	69
GLM-5.3 (low), Terminus-2	74	75	59	23	77	53	66
GLM-5.2 (Non-reasoning), Terminus-2	72	65	58	31	79	45	63
GLM-5.3-Flash (max), Terminus-2	67	86	51	38	67	48	63
GLM-5.3-Flash (low), Terminus-2	67	75	59	38	67	47	62
Claude Haiku 4.5 (Reasoning), Claude Code	65	67	49	23	56	53	58
Gemma 4 26B A4B (Reasoning), Terminus-2	59	57	47	31	70	63	56
gpt-oss-120b (high), Terminus-2	44	41	53	15	40	58	46
GLM-4.7-Flash (Reasoning), Terminus-2	42	49	41	15	47	43	43
gpt-oss-120b (high), Codex	40	23	44	0	61	62	43
GLM-4.7-Flash (Reasoning), Claude Code	41	49	39	23	49	40	41
gpt-oss-20b (high), Terminus-2	33	26	31	0	21	67	33
gpt-oss-120b (low), Terminus-2	25	34	29	8	21	60	31
gpt-oss-20b (high), Codex	11	12	5	0	4	17	9
Llama 3.3 70B (Non-reasoning), Terminus-2	7	12	10	0	2	13	8
GPT-6 Sol (medium), Codex	87	86	83	50	100	50	77, not every task attempted yet (193 of 488 tasks attempted)
GPT-5.6 Luna (xhigh), Codex	82	86	90	33	75	50	77, not every task attempted yet (210 of 488 tasks attempted)
GLM-5.2 (max), Claude Code	80	81	72	46	84	52	73, not every task attempted yet
GPT-5.6 Luna (high), Codex	78	79	69	40	87	53	72, not every task attempted yet (469 of 488 tasks attempted)
GLM-5.3-Flash (max), Claude Code	74	79	67	78	74	65	72, not every task attempted yet (449 of 488 tasks attempted)
GPT-5.6 Terra (high), Codex	76	77	61	50	74	60	69, not every task attempted yet
Gemini 3.8 Flash (low), Terminus-2	61	0	64	62	58	100	67, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (high), Terminus-2	61	40	73	54	56	97	66, not every task attempted yet (177 of 488 tasks attempted)
GPT-5.6 Luna (medium), Codex	78	76	58	0	83	41	66, not every task attempted yet (339 of 488 tasks attempted)
DeepSeek V4 Pro (high), Codex	62	0	42	17	85	65	62, not every task attempted yet (116 of 488 tasks attempted)
GPT-5.6 Luna (low), Codex	70	72	42	0	83	50	60, not every task attempted yet (397 of 488 tasks attempted)
DeepSeek V4 Flash (high), Codex	68	75	58	33	61	32	59, not every task attempted yet
Claude Sonnet 4.6 (high), Terminus-2	58	40	73	46	75	40	58, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (low), Terminus-2	55	80	45	46	60	93	57, not every task attempted yet (177 of 488 tasks attempted)
GPT-6 Luna (medium), Codex	63	65	47	22	76	38	57, not every task attempted yet (460 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (high), Terminus-2	52	60	55	31	69	62	56, not every task attempted yet (175 of 488 tasks attempted)
GPT-6 Luna (low), Codex	55	57	49	38	71	30	51, not every task attempted yet (449 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (low), Terminus-2	39	40	45	8	54	62	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.8 Flash (high), Terminus-2	35	20	45	31	32	82	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (minimal), Terminus-2	29	40	36	0	40	65	42, not every task attempted yet (177 of 488 tasks attempted)
Qwen 3.8 27B (xhigh), Terminus-2	38	No data	36	No data	No data	No data	37, not every task attempted yet (27 of 488 tasks attempted)

* Not every task attempted yet: the score may still change. What the asterisk means.

By data type

Optical multispectral score

Share of 84 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Optical multispectral score

Optical multispectral score
Model	Developer	Optical multispectral score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	57	Still to run: 63 of 84 tasks	21 of 84 tasks (11 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	61	Complete	84 of 84 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	69	Complete	84 of 84 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	63	Complete	84 of 84 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	64	Complete	84 of 84 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	38	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	33	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	43	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	29	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	52	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	29	Still to run: 63 of 84 tasks	21 of 84 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	48	Still to run: 63 of 84 tasks	21 of 84 tasks (7 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	40	Complete	84 of 84 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42	Complete	84 of 84 tasks
GLM-5.2 (high), Terminus-2	Z.ai	63	Complete	84 of 84 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	58	Complete	84 of 84 tasks
GLM-5.2 (max), Terminus-2	Z.ai	65	Complete	84 of 84 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	49	Complete	84 of 84 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	46	Complete	84 of 84 tasks
GLM-5.3 (low), Terminus-2	Z.ai	52	Complete	84 of 84 tasks
GLM-5.3 (max), Terminus-2	Z.ai	62	Complete	84 of 84 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	25	Complete	84 of 84 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	50	Complete	84 of 84 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	31	Complete	84 of 84 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	65	Complete	84 of 84 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	68	Complete	84 of 84 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	69	Complete	84 of 84 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	7	Complete	84 of 84 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	19	Still to run: 63 of 84 tasks	21 of 84 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	43	Complete	84 of 84 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	35	Complete	84 of 84 tasks
GLM-5.2 (max), Claude Code	Z.ai	70	Still to run: 2 of 84 tasks	82 of 84 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	67	Still to run: 9 of 84 tasks	75 of 84 tasks (6 attempts not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	73	Complete	84 of 84 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	54	Still to run: 1 of 84 tasks	83 of 84 tasks (48 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	52	Still to run: 63 of 84 tasks	21 of 84 tasks (44 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	43	Complete	84 of 84 tasks
gpt-oss-20b (high), Codex	OpenAI	6	Complete	84 of 84 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	60	Complete	84 of 84 tasks (8 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	71	Complete	84 of 84 tasks (1 attempt not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	70	Complete	84 of 84 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	79	Complete	84 of 84 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	54	Complete	84 of 84 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	70	Complete	84 of 84 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	68	Complete	84 of 84 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	70	Complete	84 of 84 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	74	Complete	84 of 84 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	67	Complete	84 of 84 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	68	Complete	84 of 84 tasks
Claude Opus 5 (high), Claude Code	Anthropic	73	Complete	84 of 84 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	65	Complete	84 of 84 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	70	Complete	84 of 84 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	69	Complete	84 of 84 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	71	Complete	84 of 84 tasks
GPT-6.1 Sol (high), Codex	OpenAI	71	Complete	84 of 84 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	71	Complete	84 of 84 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	65	Complete	84 of 84 tasks
GPT-6 Sol (medium), Codex	OpenAI	87	Still to run: 54 of 84 tasks	30 of 84 tasks
GPT-6 Sol (high), Codex	OpenAI	74	Complete	84 of 84 tasks
GPT-6 Luna (medium), Codex	OpenAI	44	Complete	84 of 84 tasks
GPT-6 Luna (high), Codex	OpenAI	61	Complete	84 of 84 tasks
GPT-5.6 Sol (high), Codex	OpenAI	64	Complete	84 of 84 tasks
GPT-5.6 Terra (high), Codex	OpenAI	58	Complete	84 of 84 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	53	Still to run: 10 of 84 tasks	74 of 84 tasks
GPT-5.6 Luna (high), Codex	OpenAI	65	Complete	84 of 84 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	86	Still to run: 56 of 84 tasks	28 of 84 tasks
GPT-6 Luna (low), Codex	OpenAI	43	Complete	84 of 84 tasks
GPT-5.6 Luna (low), Codex	OpenAI	36	Complete	84 of 84 tasks

Elevation or terrain score

Share of 108 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Elevation or terrain score

Elevation or terrain score
Model	Developer	Elevation or terrain score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	85	Complete	108 of 108 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	82	Complete	108 of 108 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	81	Complete	108 of 108 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	81	Complete	108 of 108 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	66	Complete	108 of 108 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	40	Complete	108 of 108 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (high), Terminus-2	Z.ai	81	Complete	108 of 108 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	79	Complete	108 of 108 tasks
GLM-5.2 (max), Terminus-2	Z.ai	81	Complete	108 of 108 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	77	Complete	108 of 108 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	76	Complete	108 of 108 tasks
GLM-5.3 (low), Terminus-2	Z.ai	85	Complete	108 of 108 tasks
GLM-5.3 (max), Terminus-2	Z.ai	87	Complete	108 of 108 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	24	Complete	108 of 108 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	44	Complete	108 of 108 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	34	Complete	108 of 108 tasks (4 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	81	Complete	108 of 108 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	84	Complete	108 of 108 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	85	Complete	108 of 108 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8	Complete	108 of 108 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	71	Complete	108 of 108 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	48	Complete	108 of 108 tasks
GLM-5.2 (max), Claude Code	Z.ai	85	Complete	108 of 108 tasks
GLM-5.3-Flash (max), Claude Code	Z.ai	75	Complete	108 of 108 tasks
GLM-5.3 (max), Claude Code	Z.ai	84	Complete	108 of 108 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	74	Complete	108 of 108 tasks (4 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	No data	Still to run: 108 of 108 tasks	0 of 108 tasks
gpt-oss-120b (high), Codex	OpenAI	31	Complete	108 of 108 tasks (5 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	14	Complete	108 of 108 tasks (11 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	81	Complete	108 of 108 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	86	Complete	108 of 108 tasks (2 attempts not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	89	Complete	108 of 108 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	92	Complete	108 of 108 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	79	Complete	108 of 108 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	81	Complete	108 of 108 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	80	Complete	108 of 108 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	85	Complete	108 of 108 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	94	Complete	108 of 108 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	85	Complete	108 of 108 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	82	Complete	108 of 108 tasks
Claude Opus 5 (high), Claude Code	Anthropic	94	Complete	108 of 108 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	81	Complete	108 of 108 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	90	Complete	108 of 108 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	84	Complete	108 of 108 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	93	Complete	108 of 108 tasks
GPT-6.1 Sol (high), Codex	OpenAI	94	Complete	108 of 108 tasks (1 attempt not counted: provider or infrastructure error)
GPT-6.1 Sol (xhigh), Codex	OpenAI	94	Complete	108 of 108 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	80	Complete	108 of 108 tasks
GPT-6 Sol (medium), Codex	OpenAI	93	Still to run: 63 of 108 tasks	45 of 108 tasks
GPT-6 Sol (high), Codex	OpenAI	93	Complete	108 of 108 tasks
GPT-6 Luna (medium), Codex	OpenAI	67	Complete	108 of 108 tasks
GPT-6 Luna (high), Codex	OpenAI	80	Complete	108 of 108 tasks
GPT-5.6 Sol (high), Codex	OpenAI	89	Complete	108 of 108 tasks
GPT-5.6 Terra (high), Codex	OpenAI	82	Complete	108 of 108 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	87	Still to run: 32 of 108 tasks	76 of 108 tasks
GPT-5.6 Luna (high), Codex	OpenAI	85	Complete	108 of 108 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	80	Still to run: 63 of 108 tasks	45 of 108 tasks
GPT-6 Luna (low), Codex	OpenAI	61	Complete	108 of 108 tasks
GPT-5.6 Luna (low), Codex	OpenAI	81	Complete	108 of 108 tasks

Climate or gridded time series score

Share of 10 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Climate or gridded time series score

Climate or gridded time series score
Model	Developer	Climate or gridded time series score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	67	Still to run: 1 of 10 tasks	9 of 10 tasks (6 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	100	Complete	10 of 10 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	90	Complete	10 of 10 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	90	Complete	10 of 10 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	80	Complete	10 of 10 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	33	Still to run: 1 of 10 tasks	9 of 10 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	33	Still to run: 1 of 10 tasks	9 of 10 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	33	Still to run: 1 of 10 tasks	9 of 10 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	78	Still to run: 1 of 10 tasks	9 of 10 tasks (1 attempt not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	89	Still to run: 1 of 10 tasks	9 of 10 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	22	Still to run: 1 of 10 tasks	9 of 10 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	67	Still to run: 1 of 10 tasks	9 of 10 tasks (1 attempt not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	60	Complete	10 of 10 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	60	Complete	10 of 10 tasks
GLM-5.2 (high), Terminus-2	Z.ai	100	Complete	10 of 10 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	80	Complete	10 of 10 tasks
GLM-5.2 (max), Terminus-2	Z.ai	100	Complete	10 of 10 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	90	Complete	10 of 10 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	90	Complete	10 of 10 tasks
GLM-5.3 (low), Terminus-2	Z.ai	90	Complete	10 of 10 tasks
GLM-5.3 (max), Terminus-2	Z.ai	80	Complete	10 of 10 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	40	Complete	10 of 10 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	40	Complete	10 of 10 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	30	Complete	10 of 10 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	90	Complete	10 of 10 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	80	Complete	10 of 10 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	80	Complete	10 of 10 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	0	Complete	10 of 10 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	100	Still to run: 8 of 10 tasks	2 of 10 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	60	Complete	10 of 10 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	30	Complete	10 of 10 tasks
GLM-5.2 (max), Claude Code	Z.ai	90	Complete	10 of 10 tasks
GLM-5.3-Flash (max), Claude Code	Z.ai	100	Still to run: 3 of 10 tasks	7 of 10 tasks
GLM-5.3 (max), Claude Code	Z.ai	80	Complete	10 of 10 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	90	Complete	10 of 10 tasks (4 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	67	Still to run: 4 of 10 tasks	6 of 10 tasks (4 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	50	Complete	10 of 10 tasks
gpt-oss-20b (high), Codex	OpenAI	10	Complete	10 of 10 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	90	Complete	10 of 10 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	80	Complete	10 of 10 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	100	Complete	10 of 10 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	80	Complete	10 of 10 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	90	Complete	10 of 10 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	100	Complete	10 of 10 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	90	Complete	10 of 10 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	90	Complete	10 of 10 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	100	Complete	10 of 10 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	100	Complete	10 of 10 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	70	Complete	10 of 10 tasks
Claude Opus 5 (high), Claude Code	Anthropic	90	Complete	10 of 10 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	100	Complete	10 of 10 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	80	Complete	10 of 10 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	100	Complete	10 of 10 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	70	Complete	10 of 10 tasks
GPT-6.1 Sol (high), Codex	OpenAI	60	Complete	10 of 10 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	70	Complete	10 of 10 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	80	Complete	10 of 10 tasks
GPT-6 Sol (medium), Codex	OpenAI	100	Still to run: 4 of 10 tasks	6 of 10 tasks
GPT-6 Sol (high), Codex	OpenAI	70	Complete	10 of 10 tasks
GPT-6 Luna (medium), Codex	OpenAI	89	Still to run: 1 of 10 tasks	9 of 10 tasks
GPT-6 Luna (high), Codex	OpenAI	70	Complete	10 of 10 tasks
GPT-5.6 Sol (high), Codex	OpenAI	90	Complete	10 of 10 tasks
GPT-5.6 Terra (high), Codex	OpenAI	90	Complete	10 of 10 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	88	Still to run: 2 of 10 tasks	8 of 10 tasks
GPT-5.6 Luna (high), Codex	OpenAI	100	Still to run: 1 of 10 tasks	9 of 10 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	100	Still to run: 3 of 10 tasks	7 of 10 tasks
GPT-6 Luna (low), Codex	OpenAI	89	Still to run: 1 of 10 tasks	9 of 10 tasks
GPT-5.6 Luna (low), Codex	OpenAI	75	Still to run: 2 of 10 tasks	8 of 10 tasks

Other gridded data score

Share of 45 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Other gridded data score

Other gridded data score
Model	Developer	Other gridded data score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	58	Still to run: 33 of 45 tasks	12 of 45 tasks (3 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	76	Complete	45 of 45 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	76	Complete	45 of 45 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	73	Complete	45 of 45 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	78	Complete	45 of 45 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	50	Still to run: 33 of 45 tasks	12 of 45 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	25	Still to run: 33 of 45 tasks	12 of 45 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	83	Still to run: 33 of 45 tasks	12 of 45 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	75	Still to run: 33 of 45 tasks	12 of 45 tasks (1 attempt not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	67	Still to run: 33 of 45 tasks	12 of 45 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	67	Still to run: 33 of 45 tasks	12 of 45 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	83	Still to run: 33 of 45 tasks	12 of 45 tasks (4 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	60	Complete	45 of 45 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	47	Complete	45 of 45 tasks
GLM-5.2 (high), Terminus-2	Z.ai	76	Complete	45 of 45 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	60	Complete	45 of 45 tasks
GLM-5.2 (max), Terminus-2	Z.ai	71	Complete	45 of 45 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	60	Complete	45 of 45 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	60	Complete	45 of 45 tasks
GLM-5.3 (low), Terminus-2	Z.ai	67	Complete	45 of 45 tasks
GLM-5.3 (max), Terminus-2	Z.ai	69	Complete	45 of 45 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	31	Complete	45 of 45 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	44	Complete	45 of 45 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	33	Complete	45 of 45 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	78	Complete	45 of 45 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	69	Complete	45 of 45 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	78	Complete	45 of 45 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	11	Complete	45 of 45 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	100	Still to run: 41 of 45 tasks	4 of 45 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	69	Complete	45 of 45 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	44	Complete	45 of 45 tasks
GLM-5.2 (max), Claude Code	Z.ai	71	Complete	45 of 45 tasks (5 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	67	Complete	45 of 45 tasks
GLM-5.3 (max), Claude Code	Z.ai	78	Complete	45 of 45 tasks
DeepSeek V4 Flash (high), Codex	DeepSeek	62	Complete	45 of 45 tasks (14 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	64	Still to run: 34 of 45 tasks	11 of 45 tasks (7 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	53	Complete	45 of 45 tasks
gpt-oss-20b (high), Codex	OpenAI	9	Complete	45 of 45 tasks
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	67	Complete	45 of 45 tasks (7 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	69	Complete	45 of 45 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	78	Complete	45 of 45 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	82	Complete	45 of 45 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	73	Complete	45 of 45 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	76	Complete	45 of 45 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	67	Complete	45 of 45 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	73	Complete	45 of 45 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	84	Complete	45 of 45 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	73	Complete	45 of 45 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	71	Complete	45 of 45 tasks
Claude Opus 5 (high), Claude Code	Anthropic	73	Complete	45 of 45 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	73	Complete	45 of 45 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	76	Complete	45 of 45 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	71	Complete	45 of 45 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	73	Complete	45 of 45 tasks
GPT-6.1 Sol (high), Codex	OpenAI	73	Complete	45 of 45 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	73	Complete	45 of 45 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	69	Complete	45 of 45 tasks
GPT-6 Sol (medium), Codex	OpenAI	74	Still to run: 18 of 45 tasks	27 of 45 tasks
GPT-6 Sol (high), Codex	OpenAI	78	Complete	45 of 45 tasks
GPT-6 Luna (medium), Codex	OpenAI	62	Complete	45 of 45 tasks
GPT-6 Luna (high), Codex	OpenAI	71	Complete	45 of 45 tasks
GPT-5.6 Sol (high), Codex	OpenAI	76	Complete	45 of 45 tasks
GPT-5.6 Terra (high), Codex	OpenAI	69	Complete	45 of 45 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	75	Still to run: 5 of 45 tasks	40 of 45 tasks
GPT-5.6 Luna (high), Codex	OpenAI	67	Complete	45 of 45 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	87	Still to run: 22 of 45 tasks	23 of 45 tasks
GPT-6 Luna (low), Codex	OpenAI	45	Still to run: 1 of 45 tasks	44 of 45 tasks
GPT-5.6 Luna (low), Codex	OpenAI	66	Still to run: 4 of 45 tasks	41 of 45 tasks

Vector only score

Share of 186 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Vector only score

Vector only score
Model	Developer	Vector only score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	70	Still to run: 113 of 186 tasks	73 of 186 tasks (33 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	73	Complete	186 of 186 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	74	Complete	186 of 186 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	77	Complete	186 of 186 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	72	Complete	186 of 186 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	45	Still to run: 113 of 186 tasks	73 of 186 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	34	Still to run: 113 of 186 tasks	73 of 186 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	62	Still to run: 115 of 186 tasks	71 of 186 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	59	Still to run: 113 of 186 tasks	73 of 186 tasks (6 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	53	Still to run: 113 of 186 tasks	73 of 186 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	32	Still to run: 113 of 186 tasks	73 of 186 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	53	Still to run: 113 of 186 tasks	73 of 186 tasks (18 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	59	Complete	186 of 186 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	46	Complete	186 of 186 tasks
GLM-5.2 (high), Terminus-2	Z.ai	77	Complete	186 of 186 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	67	Complete	186 of 186 tasks
GLM-5.2 (max), Terminus-2	Z.ai	81	Complete	186 of 186 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	70	Complete	186 of 186 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Terminus-2	Z.ai	76	Complete	186 of 186 tasks
GLM-5.3 (low), Terminus-2	Z.ai	72	Complete	186 of 186 tasks
GLM-5.3 (max), Terminus-2	Z.ai	77	Complete	186 of 186 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	28	Complete	186 of 186 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	39	Complete	186 of 186 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	23	Complete	186 of 186 tasks (2 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	71	Complete	186 of 186 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	76	Complete	186 of 186 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	76	Complete	186 of 186 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8	Complete	186 of 186 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	No data	Still to run: 186 of 186 tasks	0 of 186 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	61	Complete	186 of 186 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	47	Complete	186 of 186 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (max), Claude Code	Z.ai	79	Complete	186 of 186 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	78	Still to run: 18 of 186 tasks	168 of 186 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	76	Complete	186 of 186 tasks (3 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	68	Still to run: 1 of 186 tasks	185 of 186 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	70	Still to run: 146 of 186 tasks	40 of 186 tasks (33 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	33	Complete	186 of 186 tasks (2 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	9	Complete	186 of 186 tasks (9 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	74	Complete	186 of 186 tasks
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	76	Complete	186 of 186 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	87	Complete	186 of 186 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	85	Complete	186 of 186 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	78	Complete	186 of 186 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	83	Complete	186 of 186 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	75	Complete	186 of 186 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	78	Complete	186 of 186 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	90	Complete	186 of 186 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	85	Complete	186 of 186 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	81	Complete	186 of 186 tasks
Claude Opus 5 (high), Claude Code	Anthropic	83	Complete	186 of 186 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	77	Complete	186 of 186 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	77	Complete	186 of 186 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	74	Complete	186 of 186 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	86	Complete	186 of 186 tasks
GPT-6.1 Sol (high), Codex	OpenAI	84	Complete	186 of 186 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	87	Complete	186 of 186 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	74	Complete	186 of 186 tasks
GPT-6 Sol (medium), Codex	OpenAI	85	Still to run: 101 of 186 tasks	85 of 186 tasks
GPT-6 Sol (high), Codex	OpenAI	82	Complete	186 of 186 tasks
GPT-6 Luna (medium), Codex	OpenAI	66	Still to run: 18 of 186 tasks	168 of 186 tasks
GPT-6 Luna (high), Codex	OpenAI	76	Complete	186 of 186 tasks
GPT-5.6 Sol (high), Codex	OpenAI	84	Complete	186 of 186 tasks
GPT-5.6 Terra (high), Codex	OpenAI	74	Still to run: 1 of 186 tasks	185 of 186 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	75	Still to run: 56 of 186 tasks	130 of 186 tasks
GPT-5.6 Luna (high), Codex	OpenAI	79	Still to run: 12 of 186 tasks	174 of 186 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	84	Still to run: 83 of 186 tasks	103 of 186 tasks
GPT-6 Luna (low), Codex	OpenAI	59	Still to run: 23 of 186 tasks	163 of 186 tasks
GPT-5.6 Luna (low), Codex	OpenAI	71	Still to run: 47 of 186 tasks	139 of 186 tasks

Score by data type (0 to 100) and the Geospatial Agent Index
Model	Optical multispectral (84 tasks)	Elevation or terrain (108 tasks)	Climate or gridded time series (10 tasks)	Other gridded data (45 tasks)	Vector only (186 tasks)	Geospatial Agent Index
Claude Opus 5.5 (xhigh), Claude Code	74	94	100	84	90	85
Claude Opus 5.5 (high), Claude Code	79	92	80	82	85	82
GPT-6.1 Sol (xhigh), Codex	71	94	70	73	87	82
GPT-6.1 Sol (medium), Codex	71	93	70	73	86	82
Claude Opus 5.5 (medium), Claude Code	70	89	100	78	87	80
GPT-6.1 Sol (high), Codex	71	94	60	73	84	80
Claude Opus 5 (high), Claude Code	73	94	90	73	83	80
GPT-6 Sol (high), Codex	74	93	70	78	82	80
Claude Sonnet 5.5 (xhigh), Claude Code	67	85	100	73	85	79
Claude Sonnet 5.5 (high), Claude Code	70	81	100	76	83	78
Claude Opus 4.8 (high), Claude Code	70	90	80	76	77	76
Claude Haiku 5.5 (xhigh), Claude Code	68	82	70	71	81	76
Claude Haiku 5.5 (high), Claude Code	70	85	90	73	78	76
GPT-5.6 Sol (high), Codex	64	89	90	76	84	75
Claude Sonnet 5 (high), Claude Code	65	81	100	73	77	74
Claude Opus 4.7 (high), Claude Code	69	84	100	71	74	74
GLM-5.3 (max), Claude Code	73	84	80	78	76	73
GPT-6 Luna (xhigh), Codex	65	80	80	69	74	73
Claude Haiku 5.5 (medium), Claude Code	68	80	90	67	75	72
Kimi K3 (Reasoning), Terminus-2	69	85	80	78	76	72
DeepSeek V4 Flash (high), Terminus-2	69	82	90	76	74	72
Kimi K2.7 Code (Reasoning), Kimi Code	71	86	80	69	76	72
GLM-5.2 (max), Terminus-2	65	81	100	71	81	72
GPT-6 Luna (high), Codex	61	80	70	71	76	71
Kimi K2.7 Code (Reasoning), Terminus-2	68	84	80	69	76	71
DeepSeek V4 Pro (max), Terminus-2	63	81	90	73	77	71
Claude Sonnet 5.5 (medium), Claude Code	54	79	90	73	78	71
GLM-5.2 (high), Terminus-2	63	81	100	76	77	71
GLM-5.3 (max), Terminus-2	62	87	80	69	77	70
DeepSeek V4 Pro (high), Terminus-2	64	81	80	78	72	70
Kimi K2.6 (Reasoning), Terminus-2	65	81	90	78	71	70
DeepSeek V4 Flash (max), Terminus-2	61	85	100	76	73	69
Kimi K2.6 (Reasoning), Kimi Code	60	81	90	67	74	69
GLM-5.3 (low), Terminus-2	52	85	90	67	72	66
GLM-5.2 (Non-reasoning), Terminus-2	58	79	80	60	67	63
GLM-5.3-Flash (max), Terminus-2	46	76	90	60	76	63
GLM-5.3-Flash (low), Terminus-2	49	77	90	60	70	62
Claude Haiku 4.5 (Reasoning), Claude Code	43	71	60	69	61	58
Gemma 4 26B A4B (Reasoning), Terminus-2	40	66	60	60	59	56
gpt-oss-120b (high), Terminus-2	50	44	40	44	39	46
GLM-4.7-Flash (Reasoning), Terminus-2	42	40	60	47	46	43
gpt-oss-120b (high), Codex	43	31	50	53	33	43
GLM-4.7-Flash (Reasoning), Claude Code	35	48	30	44	47	41
gpt-oss-20b (high), Terminus-2	31	34	30	33	23	33
gpt-oss-120b (low), Terminus-2	25	24	40	31	28	31
gpt-oss-20b (high), Codex	6	14	10	9	9	9
Llama 3.3 70B (Non-reasoning), Terminus-2	7	8	0	11	8	8
GPT-6 Sol (medium), Codex	87	93	100	74	85	77, not every task attempted yet (193 of 488 tasks attempted)
GPT-5.6 Luna (xhigh), Codex	86	80	100	87	84	77, not every task attempted yet (210 of 488 tasks attempted)
GLM-5.2 (max), Claude Code	70	85	90	71	79	73, not every task attempted yet
GPT-5.6 Luna (high), Codex	65	85	100	67	79	72, not every task attempted yet (469 of 488 tasks attempted)
GLM-5.3-Flash (max), Claude Code	67	75	100	67	78	72, not every task attempted yet (449 of 488 tasks attempted)
GPT-5.6 Terra (high), Codex	58	82	90	69	74	69, not every task attempted yet
Gemini 3.8 Flash (low), Terminus-2	48	No data	67	83	53	67, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (high), Terminus-2	52	No data	89	67	53	66, not every task attempted yet (177 of 488 tasks attempted)
GPT-5.6 Luna (medium), Codex	53	87	88	75	75	66, not every task attempted yet (339 of 488 tasks attempted)
DeepSeek V4 Pro (high), Codex	52	No data	67	64	70	62, not every task attempted yet (116 of 488 tasks attempted)
GPT-5.6 Luna (low), Codex	36	81	75	66	71	60, not every task attempted yet (397 of 488 tasks attempted)
DeepSeek V4 Flash (high), Codex	54	74	90	62	68	59, not every task attempted yet
Claude Sonnet 4.6 (high), Terminus-2	57	No data	67	58	70	58, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (low), Terminus-2	29	No data	78	75	59	57, not every task attempted yet (177 of 488 tasks attempted)
GPT-6 Luna (medium), Codex	44	67	89	62	66	57, not every task attempted yet (460 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (high), Terminus-2	43	No data	33	83	62	56, not every task attempted yet (175 of 488 tasks attempted)
GPT-6 Luna (low), Codex	43	61	89	45	59	51, not every task attempted yet (449 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (low), Terminus-2	38	No data	33	50	45	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.8 Flash (high), Terminus-2	29	No data	22	67	32	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (minimal), Terminus-2	33	No data	33	25	34	42, not every task attempted yet (177 of 488 tasks attempted)
Qwen 3.8 27B (xhigh), Terminus-2	19	No data	100	100	No data	37, not every task attempted yet (27 of 488 tasks attempted)

* Not every task attempted yet: the score may still change. What the asterisk means.

A task counts under every data type it reads.

By task length

Short tasks score

Share of 177 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Short tasks score

Short tasks score
Model	Developer	Short tasks score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	68	Still to run: 133 of 177 tasks	44 of 177 tasks (19 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	88	Complete	177 of 177 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	89	Complete	177 of 177 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	86	Complete	177 of 177 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	88	Complete	177 of 177 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	91	Still to run: 133 of 177 tasks	44 of 177 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	93	Still to run: 133 of 177 tasks	44 of 177 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	86	Still to run: 133 of 177 tasks	44 of 177 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	100	Still to run: 133 of 177 tasks	44 of 177 tasks (4 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	100	Still to run: 133 of 177 tasks	44 of 177 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	91	Still to run: 133 of 177 tasks	44 of 177 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	100	Still to run: 133 of 177 tasks	44 of 177 tasks (8 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	80	Complete	177 of 177 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	66	Complete	177 of 177 tasks
GLM-5.2 (high), Terminus-2	Z.ai	86	Complete	177 of 177 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	78	Complete	177 of 177 tasks
GLM-5.2 (max), Terminus-2	Z.ai	88	Complete	177 of 177 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	81	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Terminus-2	Z.ai	89	Complete	177 of 177 tasks
GLM-5.3 (low), Terminus-2	Z.ai	85	Complete	177 of 177 tasks
GLM-5.3 (max), Terminus-2	Z.ai	87	Complete	177 of 177 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	57	Complete	177 of 177 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	59	Complete	177 of 177 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	53	Complete	177 of 177 tasks
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	88	Complete	177 of 177 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	92	Complete	177 of 177 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	88	Complete	177 of 177 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	19	Complete	177 of 177 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	100	Still to run: 169 of 177 tasks	8 of 177 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	81	Complete	177 of 177 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	64	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (max), Claude Code	Z.ai	88	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	87	Still to run: 2 of 177 tasks	175 of 177 tasks
GLM-5.3 (max), Claude Code	Z.ai	91	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	81	Complete	177 of 177 tasks (24 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	84	Still to run: 139 of 177 tasks	38 of 177 tasks (12 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	49	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	15	Complete	177 of 177 tasks (11 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	90	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	88	Complete	177 of 177 tasks (1 attempt not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	97	Complete	177 of 177 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	95	Complete	177 of 177 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	88	Complete	177 of 177 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	92	Complete	177 of 177 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	91	Complete	177 of 177 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	93	Complete	177 of 177 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	97	Complete	177 of 177 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	95	Complete	177 of 177 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	93	Complete	177 of 177 tasks
Claude Opus 5 (high), Claude Code	Anthropic	97	Complete	177 of 177 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	86	Complete	177 of 177 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	92	Complete	177 of 177 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	90	Complete	177 of 177 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	97	Complete	177 of 177 tasks
GPT-6.1 Sol (high), Codex	OpenAI	97	Complete	177 of 177 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	97	Complete	177 of 177 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	91	Complete	177 of 177 tasks
GPT-6 Sol (medium), Codex	OpenAI	93	Still to run: 73 of 177 tasks	104 of 177 tasks
GPT-6 Sol (high), Codex	OpenAI	94	Complete	177 of 177 tasks
GPT-6 Luna (medium), Codex	OpenAI	73	Still to run: 6 of 177 tasks	171 of 177 tasks
GPT-6 Luna (high), Codex	OpenAI	90	Complete	177 of 177 tasks
GPT-5.6 Sol (high), Codex	OpenAI	94	Complete	177 of 177 tasks
GPT-5.6 Terra (high), Codex	OpenAI	89	Complete	177 of 177 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	87	Still to run: 33 of 177 tasks	144 of 177 tasks
GPT-5.6 Luna (high), Codex	OpenAI	85	Still to run: 3 of 177 tasks	174 of 177 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	90	Still to run: 62 of 177 tasks	115 of 177 tasks
GPT-6 Luna (low), Codex	OpenAI	67	Still to run: 8 of 177 tasks	169 of 177 tasks
GPT-5.6 Luna (low), Codex	OpenAI	84	Still to run: 23 of 177 tasks	154 of 177 tasks

Medium tasks score

Share of 144 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Medium tasks score

Medium tasks score
Model	Developer	Medium tasks score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	58	Still to run: 99 of 144 tasks	45 of 144 tasks (20 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	75	Complete	144 of 144 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	76	Complete	144 of 144 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	72	Complete	144 of 144 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	75	Complete	144 of 144 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	49	Still to run: 99 of 144 tasks	45 of 144 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	38	Still to run: 99 of 144 tasks	45 of 144 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	71	Still to run: 99 of 144 tasks	45 of 144 tasks
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	78	Still to run: 99 of 144 tasks	45 of 144 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	73	Still to run: 99 of 144 tasks	45 of 144 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	53	Still to run: 99 of 144 tasks	45 of 144 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	76	Still to run: 99 of 144 tasks	45 of 144 tasks (14 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	58	Complete	144 of 144 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	40	Complete	144 of 144 tasks
GLM-5.2 (high), Terminus-2	Z.ai	77	Complete	144 of 144 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	71	Complete	144 of 144 tasks
GLM-5.2 (max), Terminus-2	Z.ai	77	Complete	144 of 144 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	71	Complete	144 of 144 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	73	Complete	144 of 144 tasks
GLM-5.3 (low), Terminus-2	Z.ai	74	Complete	144 of 144 tasks
GLM-5.3 (max), Terminus-2	Z.ai	81	Complete	144 of 144 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	21	Complete	144 of 144 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	42	Complete	144 of 144 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	29	Complete	144 of 144 tasks (3 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	73	Complete	144 of 144 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	72	Complete	144 of 144 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	76	Complete	144 of 144 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	3	Complete	144 of 144 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	50	Still to run: 142 of 144 tasks	2 of 144 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	59	Complete	144 of 144 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	44	Complete	144 of 144 tasks
GLM-5.2 (max), Claude Code	Z.ai	80	Complete	144 of 144 tasks
GLM-5.3-Flash (max), Claude Code	Z.ai	72	Still to run: 7 of 144 tasks	137 of 144 tasks
GLM-5.3 (max), Claude Code	Z.ai	76	Complete	144 of 144 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	69	Complete	144 of 144 tasks (31 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	73	Still to run: 118 of 144 tasks	26 of 144 tasks (27 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	42	Complete	144 of 144 tasks (2 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	10	Complete	144 of 144 tasks (7 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	74	Complete	144 of 144 tasks (5 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	75	Complete	144 of 144 tasks
Claude Opus 5.5 (medium), Claude Code	Anthropic	88	Complete	144 of 144 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	90	Complete	144 of 144 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	78	Complete	144 of 144 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	88	Complete	144 of 144 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	78	Complete	144 of 144 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	81	Complete	144 of 144 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	93	Complete	144 of 144 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	87	Complete	144 of 144 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	82	Complete	144 of 144 tasks
Claude Opus 5 (high), Claude Code	Anthropic	88	Complete	144 of 144 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	83	Complete	144 of 144 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	83	Complete	144 of 144 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	79	Complete	144 of 144 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	85	Complete	144 of 144 tasks
GPT-6.1 Sol (high), Codex	OpenAI	87	Complete	144 of 144 tasks (1 attempt not counted: provider or infrastructure error)
GPT-6.1 Sol (xhigh), Codex	OpenAI	88	Complete	144 of 144 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	76	Complete	144 of 144 tasks
GPT-6 Sol (medium), Codex	OpenAI	87	Still to run: 84 of 144 tasks	60 of 144 tasks
GPT-6 Sol (high), Codex	OpenAI	85	Complete	144 of 144 tasks
GPT-6 Luna (medium), Codex	OpenAI	65	Still to run: 12 of 144 tasks	132 of 144 tasks
GPT-6 Luna (high), Codex	OpenAI	76	Complete	144 of 144 tasks
GPT-5.6 Sol (high), Codex	OpenAI	84	Complete	144 of 144 tasks
GPT-5.6 Terra (high), Codex	OpenAI	74	Still to run: 1 of 144 tasks	143 of 144 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	75	Still to run: 42 of 144 tasks	102 of 144 tasks
GPT-5.6 Luna (high), Codex	OpenAI	82	Still to run: 7 of 144 tasks	137 of 144 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	82	Still to run: 76 of 144 tasks	68 of 144 tasks
GPT-6 Luna (low), Codex	OpenAI	52	Still to run: 14 of 144 tasks	130 of 144 tasks
GPT-5.6 Luna (low), Codex	OpenAI	71	Still to run: 24 of 144 tasks	120 of 144 tasks

Long tasks score

Share of 167 tasks passed, 0 to 100 · Higher is better

Striped: not every task attempted yet

Data table: Long tasks score

Long tasks score
Model	Developer	Long tasks score	Status	Tasks attempted
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	51	Still to run: 79 of 167 tasks	88 of 167 tasks (40 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	52	Complete	167 of 167 tasks
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	56	Complete	167 of 167 tasks
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	60	Complete	167 of 167 tasks
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	53	Complete	167 of 167 tasks
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	30	Still to run: 79 of 167 tasks	88 of 167 tasks
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	22	Still to run: 79 of 167 tasks	88 of 167 tasks
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	40	Still to run: 81 of 167 tasks	86 of 167 tasks (2 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	49	Still to run: 79 of 167 tasks	88 of 167 tasks (8 attempts not counted: provider or infrastructure error)
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	56	Still to run: 79 of 167 tasks	88 of 167 tasks
Gemini 3.8 Flash (high), Terminus-2	Google	27	Still to run: 79 of 167 tasks	88 of 167 tasks
Gemini 3.8 Flash (low), Terminus-2	Google	56	Still to run: 79 of 167 tasks	88 of 167 tasks (22 attempts not counted: provider or infrastructure error)
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	35	Complete	167 of 167 tasks
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	23	Complete	167 of 167 tasks (1 attempt not counted: provider or infrastructure error)
GLM-5.2 (high), Terminus-2	Z.ai	56	Complete	167 of 167 tasks
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	47	Complete	167 of 167 tasks
GLM-5.2 (max), Terminus-2	Z.ai	56	Complete	167 of 167 tasks
GLM-5.3-Flash (low), Terminus-2	Z.ai	43	Complete	167 of 167 tasks
GLM-5.3-Flash (max), Terminus-2	Z.ai	37	Complete	167 of 167 tasks
GLM-5.3 (low), Terminus-2	Z.ai	47	Complete	167 of 167 tasks
GLM-5.3 (max), Terminus-2	Z.ai	52	Complete	167 of 167 tasks
gpt-oss-120b (low), Terminus-2	OpenAI	12	Complete	167 of 167 tasks
gpt-oss-120b (high), Terminus-2	OpenAI	32	Complete	167 of 167 tasks
gpt-oss-20b (high), Terminus-2	OpenAI	16	Complete	167 of 167 tasks (3 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	52	Complete	167 of 167 tasks
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	55	Complete	167 of 167 tasks
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	57	Complete	167 of 167 tasks
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2	Complete	167 of 167 tasks
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	6	Still to run: 150 of 167 tasks	17 of 167 tasks
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	38	Complete	167 of 167 tasks
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	21	Complete	167 of 167 tasks
GLM-5.2 (max), Claude Code	Z.ai	58	Still to run: 2 of 167 tasks	165 of 167 tasks (10 attempts not counted: provider or infrastructure error)
GLM-5.3-Flash (max), Claude Code	Z.ai	57	Still to run: 30 of 167 tasks	137 of 167 tasks (8 attempts not counted: provider or infrastructure error)
GLM-5.3 (max), Claude Code	Z.ai	58	Complete	167 of 167 tasks (2 attempts not counted: provider or infrastructure error)
DeepSeek V4 Flash (high), Codex	DeepSeek	37	Still to run: 2 of 167 tasks	165 of 167 tasks (75 attempts not counted: provider or infrastructure error)
DeepSeek V4 Pro (high), Codex	DeepSeek	44	Still to run: 115 of 167 tasks	52 of 167 tasks (70 attempts not counted: provider or infrastructure error)
gpt-oss-120b (high), Codex	OpenAI	30	Complete	167 of 167 tasks (4 attempts not counted: provider or infrastructure error)
gpt-oss-20b (high), Codex	OpenAI	4	Complete	167 of 167 tasks (2 attempts not counted: provider or infrastructure error)
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	49	Complete	167 of 167 tasks (8 attempts not counted: provider or infrastructure error)
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	58	Complete	167 of 167 tasks (2 attempts not counted: provider or infrastructure error)
Claude Opus 5.5 (medium), Claude Code	Anthropic	63	Complete	167 of 167 tasks
Claude Opus 5.5 (high), Claude Code	Anthropic	67	Complete	167 of 167 tasks
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	54	Complete	167 of 167 tasks
Claude Sonnet 5.5 (high), Claude Code	Anthropic	61	Complete	167 of 167 tasks
Claude Haiku 5.5 (medium), Claude Code	Anthropic	53	Complete	167 of 167 tasks
Claude Haiku 5.5 (high), Claude Code	Anthropic	58	Complete	167 of 167 tasks
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	70	Complete	167 of 167 tasks
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	61	Complete	167 of 167 tasks
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	59	Complete	167 of 167 tasks
Claude Opus 5 (high), Claude Code	Anthropic	61	Complete	167 of 167 tasks
Claude Sonnet 5 (high), Claude Code	Anthropic	58	Complete	167 of 167 tasks
Claude Opus 4.8 (high), Claude Code	Anthropic	58	Complete	167 of 167 tasks
Claude Opus 4.7 (high), Claude Code	Anthropic	56	Complete	167 of 167 tasks
GPT-6.1 Sol (medium), Codex	OpenAI	69	Complete	167 of 167 tasks
GPT-6.1 Sol (high), Codex	OpenAI	65	Complete	167 of 167 tasks
GPT-6.1 Sol (xhigh), Codex	OpenAI	68	Complete	167 of 167 tasks
GPT-6 Luna (xhigh), Codex	OpenAI	56	Complete	167 of 167 tasks
GPT-6 Sol (medium), Codex	OpenAI	52	Still to run: 138 of 167 tasks	29 of 167 tasks
GPT-6 Sol (high), Codex	OpenAI	65	Complete	167 of 167 tasks
GPT-6 Luna (medium), Codex	OpenAI	40	Still to run: 10 of 167 tasks	157 of 167 tasks
GPT-6 Luna (high), Codex	OpenAI	54	Complete	167 of 167 tasks
GPT-5.6 Sol (high), Codex	OpenAI	56	Complete	167 of 167 tasks
GPT-5.6 Terra (high), Codex	OpenAI	52	Complete	167 of 167 tasks
GPT-5.6 Luna (medium), Codex	OpenAI	43	Still to run: 74 of 167 tasks	93 of 167 tasks
GPT-5.6 Luna (high), Codex	OpenAI	56	Still to run: 9 of 167 tasks	158 of 167 tasks
GPT-5.6 Luna (xhigh), Codex	OpenAI	52	Still to run: 140 of 167 tasks	27 of 167 tasks
GPT-6 Luna (low), Codex	OpenAI	38	Still to run: 17 of 167 tasks	150 of 167 tasks
GPT-5.6 Luna (low), Codex	OpenAI	35	Still to run: 44 of 167 tasks	123 of 167 tasks

Score by task length (0 to 100) and the Geospatial Agent Index
Model	Short tasks (177 tasks)	Medium tasks (144 tasks)	Long tasks (167 tasks)	Geospatial Agent Index
Claude Opus 5.5 (xhigh), Claude Code	97	93	70	85
Claude Opus 5.5 (high), Claude Code	95	90	67	82
GPT-6.1 Sol (xhigh), Codex	97	88	68	82
GPT-6.1 Sol (medium), Codex	97	85	69	82
Claude Opus 5.5 (medium), Claude Code	97	88	63	80
GPT-6.1 Sol (high), Codex	97	87	65	80
Claude Opus 5 (high), Claude Code	97	88	61	80
GPT-6 Sol (high), Codex	94	85	65	80
Claude Sonnet 5.5 (xhigh), Claude Code	95	87	61	79
Claude Sonnet 5.5 (high), Claude Code	92	88	61	78
Claude Opus 4.8 (high), Claude Code	92	83	58	76
Claude Haiku 5.5 (xhigh), Claude Code	93	82	59	76
Claude Haiku 5.5 (high), Claude Code	93	81	58	76
GPT-5.6 Sol (high), Codex	94	84	56	75
Claude Sonnet 5 (high), Claude Code	86	83	58	74
Claude Opus 4.7 (high), Claude Code	90	79	56	74
GLM-5.3 (max), Claude Code	91	76	58	73
GPT-6 Luna (xhigh), Codex	91	76	56	73
Claude Haiku 5.5 (medium), Claude Code	91	78	53	72
Kimi K3 (Reasoning), Terminus-2	88	76	57	72
DeepSeek V4 Flash (high), Terminus-2	89	76	56	72
Kimi K2.7 Code (Reasoning), Kimi Code	88	75	58	72
GLM-5.2 (max), Terminus-2	88	77	56	72
GPT-6 Luna (high), Codex	90	76	54	71
Kimi K2.7 Code (Reasoning), Terminus-2	92	72	55	71
DeepSeek V4 Pro (max), Terminus-2	86	72	60	71
Claude Sonnet 5.5 (medium), Claude Code	88	78	54	71
GLM-5.2 (high), Terminus-2	86	77	56	71
GLM-5.3 (max), Terminus-2	87	81	52	70
DeepSeek V4 Pro (high), Terminus-2	88	75	53	70
Kimi K2.6 (Reasoning), Terminus-2	88	73	52	70
DeepSeek V4 Flash (max), Terminus-2	88	75	52	69
Kimi K2.6 (Reasoning), Kimi Code	90	74	49	69
GLM-5.3 (low), Terminus-2	85	74	47	66
GLM-5.2 (Non-reasoning), Terminus-2	78	71	47	63
GLM-5.3-Flash (max), Terminus-2	89	73	37	63
GLM-5.3-Flash (low), Terminus-2	81	71	43	62
Claude Haiku 4.5 (Reasoning), Claude Code	81	59	38	58
Gemma 4 26B A4B (Reasoning), Terminus-2	80	58	35	56
gpt-oss-120b (high), Terminus-2	59	42	32	46
GLM-4.7-Flash (Reasoning), Terminus-2	66	40	23	43
gpt-oss-120b (high), Codex	49	42	30	43
GLM-4.7-Flash (Reasoning), Claude Code	64	44	21	41
gpt-oss-20b (high), Terminus-2	53	29	16	33
gpt-oss-120b (low), Terminus-2	57	21	12	31
gpt-oss-20b (high), Codex	15	10	4	9
Llama 3.3 70B (Non-reasoning), Terminus-2	19	3	2	8
GPT-6 Sol (medium), Codex	93	87	52	77, not every task attempted yet (193 of 488 tasks attempted)
GPT-5.6 Luna (xhigh), Codex	90	82	52	77, not every task attempted yet (210 of 488 tasks attempted)
GLM-5.2 (max), Claude Code	88	80	58	73, not every task attempted yet
GPT-5.6 Luna (high), Codex	85	82	56	72, not every task attempted yet (469 of 488 tasks attempted)
GLM-5.3-Flash (max), Claude Code	87	72	57	72, not every task attempted yet (449 of 488 tasks attempted)
GPT-5.6 Terra (high), Codex	89	74	52	69, not every task attempted yet
Gemini 3.8 Flash (low), Terminus-2	100	76	56	67, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (high), Terminus-2	100	73	56	66, not every task attempted yet (177 of 488 tasks attempted)
GPT-5.6 Luna (medium), Codex	87	75	43	66, not every task attempted yet (339 of 488 tasks attempted)
DeepSeek V4 Pro (high), Codex	84	73	44	62, not every task attempted yet (116 of 488 tasks attempted)
GPT-5.6 Luna (low), Codex	84	71	35	60, not every task attempted yet (397 of 488 tasks attempted)
DeepSeek V4 Flash (high), Codex	81	69	37	59, not every task attempted yet
Claude Sonnet 4.6 (high), Terminus-2	68	58	51	58, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Pro (preview) (low), Terminus-2	100	78	49	57, not every task attempted yet (177 of 488 tasks attempted)
GPT-6 Luna (medium), Codex	73	65	40	57, not every task attempted yet (460 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (high), Terminus-2	86	71	40	56, not every task attempted yet (175 of 488 tasks attempted)
GPT-6 Luna (low), Codex	67	52	38	51, not every task attempted yet (449 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (low), Terminus-2	91	49	30	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.8 Flash (high), Terminus-2	91	53	27	48, not every task attempted yet (177 of 488 tasks attempted)
Gemini 3.1 Flash-Lite (minimal), Terminus-2	93	38	22	42, not every task attempted yet (177 of 488 tasks attempted)
Qwen 3.8 27B (xhigh), Terminus-2	100	50	6	37, not every task attempted yet (27 of 488 tasks attempted)

* Not every task attempted yet: the score may still change. What the asterisk means.

Median agent time per task: short up to 65 seconds, medium up to 2.2 minutes, long more than 2.2 minutes.

The same attempts as the index, grouped by task format, kind of work, data type and task length; these scores do not enter the index. See the methodology for how each task is assigned.

By task format and kind of work

Scores

By task format

Multi-step analysis score

Single-answer questions score

Recognising infeasible requests score

By kind of work

Raster and remote sensing score

Vector and overlay score

Climate and time series score

Spatial statistics and interpolation score

Mapping and cartography score

Recognising infeasible requests score

By data type

Optical multispectral score

Elevation or terrain score

Climate or gridded time series score

Other gridded data score

Vector only score

By task length

Short tasks score

Medium tasks score

Long tasks score

Explore kinds of work