Claude Sonnet 5.5 (low)

In Claude Code · Released 28 September 2026 · Proprietary

Overview

Geospatial Intelligence Index78.2Text only · Higher is better Geospatial Agent Index69.2Text only · Higher is better Cost per task$0.043Agent tasks, at list prices · Lower is better Time per task0.4 minAt 100 output tokens/s · Lower is better

Claude Sonnet 5.5 (low) ranks 32nd of 58 models on the Geospatial Intelligence Index (Text only), at 78.2, 0.3 points below the median model. In Claude Code it ranks 41st of 74 entries on the Geospatial Agent Index, at 69.2. It costs $0.043 per task, less than the median entry ($0.081) and takes 0.4 min per task, faster than the median entry (2.0 min).

Compare head to head: GPT-5.6 Terra (high), Codex · DeepSeek V4 Flash (max), Terminus-2 · Kimi K2.6 (Reasoning), Kimi Code · Kimi K2.6 (Reasoning), Terminus-2 · Claude Haiku 5.5 (low), Claude Code · Qwen 3.8 27B (xhigh), Qwen Code · all comparisons.

Technical specification

Developer: Anthropic
Released: 28 September 2026
Weights: Not published
Harness: Claude Code 2.1.286
Reasoning setting: low
Context window: 1,000,000 tokens
Maximum output tokens: 128,000
Image input: Yes
Temperature: Provider default
Input price per 1M tokens: $2.00
Cached input price per 1M tokens: $0.10
Output price per 1M tokens: $10.00
Map and trace judge: Claude Sonnet 5.5
Reward hacking: 5.7% of reviewed passing attempts flaggedSafety page

How models are selected.

Intelligence

All

Geospatial Intelligence Index: Text only

Weighted average of agent tasks, guided workflows and geospatial knowledge, 0 to 100 · Higher is better

How to read this chart

What this metric means. A weighted average of group scores, each 0 to 100: Agent tasks 4/6, Guided workflows 1/6, Geospatial knowledge 1/6.

Data table: Geospatial Intelligence Index: Text only

Geospatial Intelligence Index: Text only
Model	Developer	Geospatial Intelligence Index: Text only
Claude Opus 5.5 (xhigh)	Anthropic	88.7
GPT-6.1 Sol (xhigh)	OpenAI	86.2
GPT-6 Sol (high)	OpenAI	84.9
Claude Opus 5 (high)	Anthropic	84.5
Claude Sonnet 5.5 (xhigh)	Anthropic	84.4
Claude Opus 4.8 (high)	Anthropic	82.5
Claude Haiku 5.5 (xhigh)	Anthropic	82.0
GPT-5.6 Sol (high)	OpenAI	81.3
Claude Sonnet 5 (high)	Anthropic	80.9
Claude Opus 4.7 (high)	Anthropic	80.8
GPT-5.6 Luna (xhigh)	OpenAI	80.0
GPT-6 Luna (xhigh)	OpenAI	80.0
DeepSeek V4 Pro (max) Open weights	DeepSeek	79.0
Gemini 3.8 Flash (low)	Google	78.7
Kimi K3 (Reasoning) Open weights	Moonshot AI	78.6
GLM-5.2 (max) Open weights	Z.ai	78.4
Claude Sonnet 5.5 (low)	Anthropic	78.2
GPT-5.6 Terra (high)	OpenAI	77.5
DeepSeek V4 Flash (max) Open weights	DeepSeek	77.2
GLM-5.3 (max) Open weights	Z.ai	77.1
Qwen 3.8 27B (xhigh) Open weights	Alibaba (Qwen)	75.1
Nemotron 3 Ultra (high) Open weights	NVIDIA	73.5
GLM-5.3-Flash (max) Open weights	Z.ai	71.0
Claude Haiku 4.5 (Reasoning)	Anthropic	65.8
Gemma 4 26B A4B (Reasoning) Open weights	Google	64.4
Gemini 3.1 Flash-Lite (high)	Google	63.2
Nemotron 3.5 Super VL 120B (Reasoning) Open weights	NVIDIA	58.5
gpt-oss-120b (high) Open weights	OpenAI	51.9
GLM-4.7-Flash (Reasoning) Open weights	Z.ai	49.6
Mistral Large 3 (Non-reasoning) Open weights	Mistral AI	44.5
Nemotron 3 Nano Omni 30B (Reasoning) Open weights	NVIDIA	31.1
gpt-oss-20b (high) Open weights	OpenAI	22.1
Llama 3.3 70B (Non-reasoning) Open weights	Meta	19.9

Open weights

Geospatial Intelligence Index: Text only

Weighted average of agent tasks, guided workflows and geospatial knowledge, 0 to 100 · Higher is better

How to read this chart

What this metric means. A weighted average of group scores, each 0 to 100: Agent tasks 4/6, Guided workflows 1/6, Geospatial knowledge 1/6.

Data table: Geospatial Intelligence Index: Text only

Geospatial Intelligence Index: Text only
Model	Developer	Geospatial Intelligence Index: Text only
DeepSeek V4 Pro (max) Open weights	DeepSeek	79.0
Kimi K3 (Reasoning) Open weights	Moonshot AI	78.6
GLM-5.2 (max) Open weights	Z.ai	78.4
Claude Sonnet 5.5 (low)	Anthropic	78.2
DeepSeek V4 Flash (max) Open weights	DeepSeek	77.2
GLM-5.3 (max) Open weights	Z.ai	77.1
Qwen 3.8 27B (xhigh) Open weights	Alibaba (Qwen)	75.1
Nemotron 3 Ultra (high) Open weights	NVIDIA	73.5
GLM-5.3-Flash (max) Open weights	Z.ai	71.0
Gemma 4 26B A4B (Reasoning) Open weights	Google	64.4
Nemotron 3.5 Super VL 120B (Reasoning) Open weights	NVIDIA	58.5
gpt-oss-120b (high) Open weights	OpenAI	51.9
GLM-4.7-Flash (Reasoning) Open weights	Z.ai	49.6
Mistral Large 3 (Non-reasoning) Open weights	Mistral AI	44.5
Nemotron 3 Nano Omni 30B (Reasoning) Open weights	NVIDIA	31.1
gpt-oss-20b (high) Open weights	OpenAI	22.1
Llama 3.3 70B (Non-reasoning) Open weights	Meta	19.9

Agent tasks by evaluation

All

Geospatial Agent Index: Text only

Equal average of Earth-Bench V2, GeoBenchX and GS-QA2, 0 to 100 · Higher is better

Data table: Geospatial Agent Index: Text only

Geospatial Agent Index: Text only
Model	Developer	Geospatial Agent Index: Text only
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	85
Claude Opus 5.5 (high), Claude Code	Anthropic	82
GPT-6.1 Sol (xhigh), Codex	OpenAI	82
GPT-6.1 Sol (medium), Codex	OpenAI	82
Claude Opus 5.5 (medium), Claude Code	Anthropic	80
GPT-6.1 Sol (high), Codex	OpenAI	80
Claude Opus 5 (high), Claude Code	Anthropic	80
GPT-6 Sol (high), Codex	OpenAI	80
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	79
Claude Sonnet 5.5 (high), Claude Code	Anthropic	78
Claude Opus 4.8 (high), Claude Code	Anthropic	76
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	76
Claude Haiku 5.5 (high), Claude Code	Anthropic	76
GPT-5.6 Sol (high), Codex	OpenAI	75
Claude Sonnet 5 (high), Claude Code	Anthropic	74
Claude Opus 5.5 (low), Claude Code	Anthropic	74
Claude Opus 4.7 (high), Claude Code	Anthropic	74
GLM-5.3 (max), Claude Code	Z.ai	73
GLM-5.2 (max), Claude Code	Z.ai	73
GPT-5.6 Luna (xhigh), Codex	OpenAI	73
GPT-6 Luna (xhigh), Codex	OpenAI	73
Claude Haiku 5.5 (medium), Claude Code	Anthropic	72
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	72
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	72
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	72
GLM-5.2 (max), Terminus-2	Z.ai	72
GPT-5.6 Luna (high), Codex	OpenAI	72
GLM-5.3-Flash (max), Claude Code	Z.ai	72
GPT-6 Luna (high), Codex	OpenAI	71
Gemini 3.8 Flash (low), Terminus-2	Google	71
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	71
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	71
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	71
GLM-5.2 (high), Terminus-2	Z.ai	71
DeepSeek V4.1 Flash (max), Codex	DeepSeek	71
GLM-5.3 (max), Terminus-2	Z.ai	70
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	70
Claude Haiku 5.5 (low), Claude Code	Anthropic	70
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	70
GPT-5.6 Terra (high), Codex	OpenAI	69
Claude Sonnet 5.5 (low), Claude Code	Anthropic	69
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	69
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	69
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	68
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	68
GPT-5.6 Luna (medium), Codex	OpenAI	66
GLM-5.3 (low), Terminus-2	Z.ai	66
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	63
GLM-5.3-Flash (max), Terminus-2	Z.ai	63
GLM-5.3-Flash (low), Terminus-2	Z.ai	62
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	61
GPT-5.6 Luna (low), Codex	OpenAI	61
DeepSeek V4 Flash (high), Codex	DeepSeek	59
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	58
GPT-6 Luna (medium), Codex	OpenAI	57
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	56
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	56
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	52
GPT-6 Luna (low), Codex	OpenAI	51
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	51
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	48
gpt-oss-120b (high), Terminus-2	OpenAI	46
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43
gpt-oss-120b (high), Codex	OpenAI	43
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	41
Step 5 Preview (Reasoning), Terminus-2	StepFun	34
gpt-oss-20b (high), Terminus-2	OpenAI	33
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	33
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	32
gpt-oss-120b (low), Terminus-2	OpenAI	31
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	20
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	11
gpt-oss-20b (high), Codex	OpenAI	9
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8

Open weights

Geospatial Agent Index: Text only

Equal average of Earth-Bench V2, GeoBenchX and GS-QA2, 0 to 100 · Higher is better

Data table: Geospatial Agent Index: Text only

Geospatial Agent Index: Text only
Model	Developer	Geospatial Agent Index: Text only
GLM-5.3 (max), Claude Code	Z.ai	73
GLM-5.2 (max), Claude Code	Z.ai	73
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	72
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	72
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	72
GLM-5.2 (max), Terminus-2	Z.ai	72
GLM-5.3-Flash (max), Claude Code	Z.ai	72
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	71
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	71
GLM-5.2 (high), Terminus-2	Z.ai	71
DeepSeek V4.1 Flash (max), Codex	DeepSeek	71
GLM-5.3 (max), Terminus-2	Z.ai	70
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	70
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	70
Claude Sonnet 5.5 (low), Claude Code	Anthropic	69
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	69
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	69
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	68
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	68
GLM-5.3 (low), Terminus-2	Z.ai	66
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	63
GLM-5.3-Flash (max), Terminus-2	Z.ai	63
GLM-5.3-Flash (low), Terminus-2	Z.ai	62
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	61
DeepSeek V4 Flash (high), Codex	DeepSeek	59
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	56
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	52
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	51
gpt-oss-120b (high), Terminus-2	OpenAI	46
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43
gpt-oss-120b (high), Codex	OpenAI	43
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	41
Step 5 Preview (Reasoning), Terminus-2	StepFun	34
gpt-oss-20b (high), Terminus-2	OpenAI	33
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	33
gpt-oss-120b (low), Terminus-2	OpenAI	31
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	20
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	11
gpt-oss-20b (high), Codex	OpenAI	9
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8

All

Earth-Bench V2 score

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

How to read this chart

About this evaluation. Earth observation data analysis. Earth-Bench V2 results and background.

Data table: Earth-Bench V2 score

Earth-Bench V2 score
Model	Developer	Earth-Bench V2 score
Claude Opus 5.5 (high), Claude Code	Anthropic	79
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	77
GPT-6 Sol (medium), Codex	OpenAI	77
GPT-6 Sol (high), Codex	OpenAI	75
GLM-5.3 (max), Claude Code	Z.ai	74
Claude Opus 5 (high), Claude Code	Anthropic	74
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	72
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	72
Claude Opus 5.5 (medium), Claude Code	Anthropic	72
Claude Haiku 5.5 (high), Claude Code	Anthropic	72
GPT-6.1 Sol (medium), Codex	OpenAI	72
GPT-6.1 Sol (high), Codex	OpenAI	72
GPT-6.1 Sol (xhigh), Codex	OpenAI	72
GLM-5.2 (max), Claude Code	Z.ai	71
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	71
Claude Sonnet 5.5 (high), Claude Code	Anthropic	71
Claude Opus 4.8 (high), Claude Code	Anthropic	71
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	69
Claude Opus 4.7 (high), Claude Code	Anthropic	69
GLM-5.3-Flash (max), Claude Code	Z.ai	69
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	68
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	68
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	68
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	68
Claude Sonnet 5 (high), Claude Code	Anthropic	68
GPT-5.6 Sol (high), Codex	OpenAI	68
GLM-5.2 (high), Terminus-2	Z.ai	68
GLM-5.2 (max), Terminus-2	Z.ai	68
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	68
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	68
Claude Haiku 5.5 (medium), Claude Code	Anthropic	68
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	67
GPT-6 Luna (xhigh), Codex	OpenAI	67
GPT-5.6 Luna (high), Codex	OpenAI	67
DeepSeek V4.1 Flash (max), Codex	DeepSeek	66
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	66
GPT-5.6 Luna (xhigh), Codex	OpenAI	65
GLM-5.3 (max), Terminus-2	Z.ai	64
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	63
GPT-6 Luna (high), Codex	OpenAI	63
Claude Haiku 5.5 (low), Claude Code	Anthropic	62
GPT-5.6 Terra (high), Codex	OpenAI	62
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	59
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	59
Claude Opus 5.5 (low), Claude Code	Anthropic	59
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	58
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	58
GLM-5.3 (low), Terminus-2	Z.ai	57
DeepSeek V4 Pro (high), Codex	DeepSeek	57
DeepSeek V4 Flash (high), Codex	DeepSeek	57
GPT-5.6 Luna (medium), Codex	OpenAI	56
Claude Sonnet 5.5 (low), Claude Code	Anthropic	56
Gemini 3.8 Flash (low), Terminus-2	Google	54
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	53
GLM-5.3-Flash (low), Terminus-2	Z.ai	53
GLM-5.3-Flash (max), Terminus-2	Z.ai	51
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	51
gpt-oss-120b (high), Terminus-2	OpenAI	50
GPT-6 Luna (medium), Codex	OpenAI	50
gpt-oss-120b (high), Codex	OpenAI	49
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	46
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	46
Gemini 3.8 Flash (high), Terminus-2	Google	44
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	44
GPT-5.6 Luna (low), Codex	OpenAI	44
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	44
GPT-6 Luna (low), Codex	OpenAI	44
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	43
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	41
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	39
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	38
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	34
gpt-oss-20b (high), Terminus-2	OpenAI	32
Step 5 Preview (Reasoning), Terminus-2	StepFun	32
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	31
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	29
gpt-oss-120b (low), Terminus-2	OpenAI	28
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	16
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	9
gpt-oss-20b (high), Codex	OpenAI	6

Open weights

Earth-Bench V2 score

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

How to read this chart

About this evaluation. Earth observation data analysis. Earth-Bench V2 results and background.

Data table: Earth-Bench V2 score

Earth-Bench V2 score
Model	Developer	Earth-Bench V2 score
GLM-5.3 (max), Claude Code	Z.ai	74
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	72
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	72
GLM-5.2 (max), Claude Code	Z.ai	71
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	71
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	69
GLM-5.3-Flash (max), Claude Code	Z.ai	69
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	68
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	68
GLM-5.2 (high), Terminus-2	Z.ai	68
GLM-5.2 (max), Terminus-2	Z.ai	68
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	68
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	68
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	67
DeepSeek V4.1 Flash (max), Codex	DeepSeek	66
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	66
GLM-5.3 (max), Terminus-2	Z.ai	64
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	63
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	58
GLM-5.3 (low), Terminus-2	Z.ai	57
DeepSeek V4 Pro (high), Codex	DeepSeek	57
DeepSeek V4 Flash (high), Codex	DeepSeek	57
Claude Sonnet 5.5 (low), Claude Code	Anthropic	56
GLM-5.3-Flash (low), Terminus-2	Z.ai	53
GLM-5.3-Flash (max), Terminus-2	Z.ai	51
gpt-oss-120b (high), Terminus-2	OpenAI	50
gpt-oss-120b (high), Codex	OpenAI	49
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	46
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	46
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	44
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	44
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	39
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	38
gpt-oss-20b (high), Terminus-2	OpenAI	32
Step 5 Preview (Reasoning), Terminus-2	StepFun	32
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	31
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	29
gpt-oss-120b (low), Terminus-2	OpenAI	28
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	16
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	9
gpt-oss-20b (high), Codex	OpenAI	6

All

GeoBenchX score

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

How to read this chart

About this evaluation. Maps, numbers and lists from country and city data; some tasks should be declined. GeoBenchX results and background.

Data table: GeoBenchX score

GeoBenchX score
Model	Developer	GeoBenchX score
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	83
Claude Sonnet 5.5 (high), Claude Code	Anthropic	81
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	81
Claude Opus 5.5 (medium), Claude Code	Anthropic	79
GPT-6.1 Sol (medium), Codex	OpenAI	79
GPT-6.1 Sol (xhigh), Codex	OpenAI	79
Claude Opus 5.5 (high), Claude Code	Anthropic	79
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	77
Claude Sonnet 5 (high), Claude Code	Anthropic	77
Claude Opus 5.5 (low), Claude Code	Anthropic	77
Claude Sonnet 5.5 (low), Claude Code	Anthropic	77
GPT-6.1 Sol (high), Codex	OpenAI	77
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	76
GPT-6 Sol (medium), Codex	OpenAI	75
Claude Opus 5 (high), Claude Code	Anthropic	75
GPT-6 Sol (high), Codex	OpenAI	75
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	74
Gemini 3.8 Flash (low), Terminus-2	Google	74
GPT-6 Luna (high), Codex	OpenAI	74
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	73
Claude Opus 4.8 (high), Claude Code	Anthropic	73
GPT-6 Luna (xhigh), Codex	OpenAI	73
Claude Haiku 5.5 (low), Claude Code	Anthropic	73
Claude Opus 4.7 (high), Claude Code	Anthropic	73
Claude Haiku 5.5 (high), Claude Code	Anthropic	72
Claude Haiku 5.5 (medium), Claude Code	Anthropic	71
GLM-5.3-Flash (max), Claude Code	Z.ai	70
GPT-5.6 Sol (high), Codex	OpenAI	70
GPT-5.6 Luna (xhigh), Codex	OpenAI	68
GLM-5.2 (max), Terminus-2	Z.ai	67
GPT-5.6 Luna (high), Codex	OpenAI	67
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	67
GLM-5.2 (high), Terminus-2	Z.ai	67
GPT-5.6 Terra (high), Codex	OpenAI	67
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	66
GLM-5.2 (max), Claude Code	Z.ai	66
DeepSeek V4 Pro (high), Codex	DeepSeek	66
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	65
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	65
GLM-5.3 (max), Terminus-2	Z.ai	65
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	65
GLM-5.3 (max), Claude Code	Z.ai	65
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	64
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	64
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	64
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	64
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63
GPT-5.6 Luna (medium), Codex	OpenAI	63
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	62
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	62
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	62
GPT-5.6 Luna (low), Codex	OpenAI	62
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	61
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	60
GLM-5.3 (low), Terminus-2	Z.ai	60
GPT-6 Luna (medium), Codex	OpenAI	57
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	57
GLM-5.3-Flash (low), Terminus-2	Z.ai	57
GLM-5.3-Flash (max), Terminus-2	Z.ai	57
DeepSeek V4.1 Flash (max), Codex	DeepSeek	56
gpt-oss-120b (high), Codex	OpenAI	55
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	53
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	53
GPT-6 Luna (low), Codex	OpenAI	51
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	51
Gemini 3.8 Flash (high), Terminus-2	Google	51
DeepSeek V4 Flash (high), Codex	DeepSeek	47
gpt-oss-120b (high), Terminus-2	OpenAI	45
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	44
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	41
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	39
gpt-oss-20b (high), Terminus-2	OpenAI	38
gpt-oss-120b (low), Terminus-2	OpenAI	37
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	29
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	21
Step 5 Preview (Reasoning), Terminus-2	StepFun	17
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	12
gpt-oss-20b (high), Codex	OpenAI	10
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	6

Open weights

GeoBenchX score

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

How to read this chart

About this evaluation. Maps, numbers and lists from country and city data; some tasks should be declined. GeoBenchX results and background.

Data table: GeoBenchX score

GeoBenchX score
Model	Developer	GeoBenchX score
Claude Sonnet 5.5 (low), Claude Code	Anthropic	77
GLM-5.3-Flash (max), Claude Code	Z.ai	70
GLM-5.2 (max), Terminus-2	Z.ai	67
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	67
GLM-5.2 (high), Terminus-2	Z.ai	67
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	66
GLM-5.2 (max), Claude Code	Z.ai	66
DeepSeek V4 Pro (high), Codex	DeepSeek	66
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	65
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	65
GLM-5.3 (max), Terminus-2	Z.ai	65
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	65
GLM-5.3 (max), Claude Code	Z.ai	65
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	64
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	64
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	64
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	64
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	62
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	62
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	62
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	60
GLM-5.3 (low), Terminus-2	Z.ai	60
GLM-5.3-Flash (low), Terminus-2	Z.ai	57
GLM-5.3-Flash (max), Terminus-2	Z.ai	57
DeepSeek V4.1 Flash (max), Codex	DeepSeek	56
gpt-oss-120b (high), Codex	OpenAI	55
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	53
DeepSeek V4 Flash (high), Codex	DeepSeek	47
gpt-oss-120b (high), Terminus-2	OpenAI	45
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	41
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	39
gpt-oss-20b (high), Terminus-2	OpenAI	38
gpt-oss-120b (low), Terminus-2	OpenAI	37
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	29
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	21
Step 5 Preview (Reasoning), Terminus-2	StepFun	17
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	12
gpt-oss-20b (high), Codex	OpenAI	10
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	6

All

GS-QA2 score

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

How to read this chart

About this evaluation. Spatial data reasoning over map and elevation data. GS-QA2 results and background.

Data table: GS-QA2 score

GS-QA2 score
Model	Developer	GS-QA2 score
GPT-5.6 Sol (medium), Codex	OpenAI	100
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	94
GPT-6.1 Sol (xhigh), Codex	OpenAI	94
GPT-6.1 Sol (medium), Codex	OpenAI	93
GPT-6.1 Sol (high), Codex	OpenAI	93
Claude Opus 5 (high), Claude Code	Anthropic	91
Claude Opus 5.5 (medium), Claude Code	Anthropic	90
Claude Opus 5.5 (high), Claude Code	Anthropic	90
DeepSeek V4.1 Flash (max), Codex	DeepSeek	89
GPT-6 Sol (high), Codex	OpenAI	89
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	87
GPT-6 Sol (medium), Codex	OpenAI	87
GPT-5.6 Sol (high), Codex	OpenAI	87
Claude Opus 5.5 (low), Claude Code	Anthropic	86
Gemini 3.8 Flash (low), Terminus-2	Google	86
GPT-5.6 Luna (xhigh), Codex	OpenAI	85
Claude Opus 4.8 (high), Claude Code	Anthropic	84
Claude Haiku 5.5 (high), Claude Code	Anthropic	83
Claude Sonnet 5.5 (high), Claude Code	Anthropic	83
GLM-5.2 (max), Claude Code	Z.ai	83
GLM-5.3 (max), Terminus-2	Z.ai	83
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	82
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	81
GLM-5.3 (max), Claude Code	Z.ai	81
GPT-5.6 Luna (high), Codex	OpenAI	81
GLM-5.2 (max), Terminus-2	Z.ai	80
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	80
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	80
GPT-5.6 Luna (medium), Codex	OpenAI	80
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	80
GPT-5.6 Terra (high), Codex	OpenAI	80
GLM-5.3 (low), Terminus-2	Z.ai	80
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	80
GLM-5.3-Flash (max), Terminus-2	Z.ai	80
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	80
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	79
Claude Haiku 5.5 (medium), Claude Code	Anthropic	79
Claude Opus 4.7 (high), Claude Code	Anthropic	79
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	79
GLM-5.2 (high), Terminus-2	Z.ai	78
GPT-6 Luna (xhigh), Codex	OpenAI	78
Claude Sonnet 5 (high), Claude Code	Anthropic	77
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	77
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	77
GPT-6 Luna (high), Codex	OpenAI	77
GLM-5.3-Flash (max), Claude Code	Z.ai	76
GPT-5.6 Luna (low), Codex	OpenAI	76
GLM-5.3-Flash (low), Terminus-2	Z.ai	76
Claude Sonnet 5.5 (low), Claude Code	Anthropic	76
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	75
Claude Haiku 5.5 (low), Claude Code	Anthropic	75
DeepSeek V4 Flash (high), Codex	DeepSeek	74
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	74
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	72
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	71
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	70
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	69
GPT-6 Luna (medium), Codex	OpenAI	65
GPT-6 Luna (low), Codex	OpenAI	59
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	59
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	58
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	55
Step 5 Preview (Reasoning), Terminus-2	StepFun	52
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	49
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	48
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43
gpt-oss-120b (high), Terminus-2	OpenAI	42
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	39
gpt-oss-20b (high), Terminus-2	OpenAI	30
gpt-oss-120b (low), Terminus-2	OpenAI	29
gpt-oss-120b (high), Codex	OpenAI	25
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	18
gpt-oss-20b (high), Codex	OpenAI	12
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	10
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	9
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	4

Open weights

GS-QA2 score

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

How to read this chart

About this evaluation. Spatial data reasoning over map and elevation data. GS-QA2 results and background.

Data table: GS-QA2 score

GS-QA2 score
Model	Developer	GS-QA2 score
DeepSeek V4.1 Flash (max), Codex	DeepSeek	89
GLM-5.2 (max), Claude Code	Z.ai	83
GLM-5.3 (max), Terminus-2	Z.ai	83
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	82
GLM-5.3 (max), Claude Code	Z.ai	81
GLM-5.2 (max), Terminus-2	Z.ai	80
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	80
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	80
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	80
GLM-5.3 (low), Terminus-2	Z.ai	80
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	80
GLM-5.3-Flash (max), Terminus-2	Z.ai	80
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	80
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	79
GLM-5.2 (high), Terminus-2	Z.ai	78
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	77
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	77
GLM-5.3-Flash (max), Claude Code	Z.ai	76
GLM-5.3-Flash (low), Terminus-2	Z.ai	76
Claude Sonnet 5.5 (low), Claude Code	Anthropic	76
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	75
DeepSeek V4 Flash (high), Codex	DeepSeek	74
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	74
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	72
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	71
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	70
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	59
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	58
Step 5 Preview (Reasoning), Terminus-2	StepFun	52
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	48
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	43
gpt-oss-120b (high), Terminus-2	OpenAI	42
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	39
gpt-oss-20b (high), Terminus-2	OpenAI	30
gpt-oss-120b (low), Terminus-2	OpenAI	29
gpt-oss-120b (high), Codex	OpenAI	25
gpt-oss-20b (high), Codex	OpenAI	12
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	10
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	9
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	4

All

GeoAgentBench score

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

How to read this chart

About this evaluation. Multi-step spatial execution. GeoAgentBench results and background.

Data table: GeoAgentBench score

GeoAgentBench score
Model	Developer	GeoAgentBench score
Claude Opus 5.5 (high), Claude Code	Anthropic	100
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	100
Claude Opus 5 (high), Claude Code	Anthropic	100
GPT-6.1 Sol (medium), Codex	OpenAI	100
GPT-6.1 Sol (high), Codex	OpenAI	100
GPT-6.1 Sol (xhigh), Codex	OpenAI	100
GPT-6 Sol (medium), Codex	OpenAI	100
GPT-6 Sol (high), Codex	OpenAI	100
GPT-5.6 Sol (high), Codex	OpenAI	100
Gemini 3.8 Flash (low), Terminus-2	Google	98
GLM-5.3-Flash (max), Claude Code	Z.ai	98
GLM-5.3 (max), Claude Code	Z.ai	98
Claude Opus 5.5 (medium), Claude Code	Anthropic	98
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	98
Claude Sonnet 5.5 (high), Claude Code	Anthropic	98
Claude Haiku 5.5 (high), Claude Code	Anthropic	98
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	98
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	98
Claude Sonnet 5 (high), Claude Code	Anthropic	98
GPT-5.6 Luna (medium), Codex	OpenAI	98
GPT-5.6 Luna (high), Codex	OpenAI	98
GPT-5.6 Luna (xhigh), Codex	OpenAI	98
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	96
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	96
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	96
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	96
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	96
Claude Opus 5.5 (low), Claude Code	Anthropic	96
Claude Opus 4.8 (high), Claude Code	Anthropic	96
Claude Opus 4.7 (high), Claude Code	Anthropic	96
GPT-6 Luna (xhigh), Codex	OpenAI	96
GPT-6 Luna (high), Codex	OpenAI	96
GPT-5.6 Terra (high), Codex	OpenAI	96
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	94
GLM-5.2 (high), Terminus-2	Z.ai	94
GLM-5.2 (max), Terminus-2	Z.ai	94
GLM-5.3-Flash (low), Terminus-2	Z.ai	94
GLM-5.3 (max), Terminus-2	Z.ai	94
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	94
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	94
Claude Haiku 5.5 (medium), Claude Code	Anthropic	94
Claude Sonnet 5.5 (low), Claude Code	Anthropic	94
GPT-6 Luna (medium), Codex	OpenAI	94
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	92
GLM-5.3 (low), Terminus-2	Z.ai	92
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	92
Claude Haiku 5.5 (low), Claude Code	Anthropic	92
GPT-6 Luna (low), Codex	OpenAI	92
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	90
DeepSeek V4.1 Flash (max), Codex	DeepSeek	90
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	88
DeepSeek V4 Flash (high), Codex	DeepSeek	88
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	86
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	86
GPT-5.6 Luna (low), Codex	OpenAI	86
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	84
GLM-5.3-Flash (max), Terminus-2	Z.ai	84
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	78
Gemini 3.8 Flash (high), Terminus-2	Google	74
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	70
gpt-oss-120b (high), Terminus-2	OpenAI	70
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	68
gpt-oss-120b (low), Terminus-2	OpenAI	60
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	60
gpt-oss-120b (high), Codex	OpenAI	58
gpt-oss-20b (high), Terminus-2	OpenAI	53
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	52
Step 5 Preview (Reasoning), Terminus-2	StepFun	50
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	46
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	42
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	30
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	26
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	10
gpt-oss-20b (high), Codex	OpenAI	10
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8

Open weights

GeoAgentBench score

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

How to read this chart

About this evaluation. Multi-step spatial execution. GeoAgentBench results and background.

Data table: GeoAgentBench score

GeoAgentBench score
Model	Developer	GeoAgentBench score
GLM-5.3-Flash (max), Claude Code	Z.ai	98
GLM-5.3 (max), Claude Code	Z.ai	98
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	96
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	96
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	96
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	96
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	94
GLM-5.2 (high), Terminus-2	Z.ai	94
GLM-5.2 (max), Terminus-2	Z.ai	94
GLM-5.3-Flash (low), Terminus-2	Z.ai	94
GLM-5.3 (max), Terminus-2	Z.ai	94
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	94
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	94
Claude Sonnet 5.5 (low), Claude Code	Anthropic	94
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	92
GLM-5.3 (low), Terminus-2	Z.ai	92
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	92
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	90
DeepSeek V4.1 Flash (max), Codex	DeepSeek	90
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	88
DeepSeek V4 Flash (high), Codex	DeepSeek	88
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	86
GLM-5.3-Flash (max), Terminus-2	Z.ai	84
gpt-oss-120b (high), Terminus-2	OpenAI	70
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	68
gpt-oss-120b (low), Terminus-2	OpenAI	60
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	60
gpt-oss-120b (high), Codex	OpenAI	58
gpt-oss-20b (high), Terminus-2	OpenAI	53
Step 5 Preview (Reasoning), Terminus-2	StepFun	50
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	46
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	42
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	30
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	26
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	10
gpt-oss-20b (high), Codex	OpenAI	10
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	8

All

GeoAnalystBench score

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

How to read this chart

About this evaluation. Spatial workflow and code generation. GeoAnalystBench results and background.

Data table: GeoAnalystBench score

GeoAnalystBench score
Model	Developer	GeoAnalystBench score
Claude Sonnet 4.6 (high), Terminus-2	Anthropic	100
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	100
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	100
GLM-5.2 (high), Terminus-2	Z.ai	100
GLM-5.3-Flash (max), Claude Code	Z.ai	100
Claude Opus 5.5 (high), Claude Code	Anthropic	100
Claude Sonnet 5.5 (high), Claude Code	Anthropic	100
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	100
Claude Opus 5.5 (low), Claude Code	Anthropic	100
Claude Haiku 5.5 (low), Claude Code	Anthropic	100
Claude Opus 4.8 (high), Claude Code	Anthropic	100
Claude Opus 4.7 (high), Claude Code	Anthropic	100
GPT-6.1 Sol (medium), Codex	OpenAI	100
GPT-6.1 Sol (high), Codex	OpenAI	100
GPT-6 Luna (xhigh), Codex	OpenAI	100
GPT-6 Sol (high), Codex	OpenAI	100
GPT-6 Luna (high), Codex	OpenAI	100
GPT-5.6 Luna (medium), Codex	OpenAI	100
GPT-5.6 Luna (high), Codex	OpenAI	100
GPT-5.6 Luna (xhigh), Codex	OpenAI	100
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	95
Gemini 3.1 Pro (preview) (high), Terminus-2	Google	95
GLM-5.2 (max), Terminus-2	Z.ai	95
GLM-5.3-Flash (low), Terminus-2	Z.ai	95
GLM-5.3 (low), Terminus-2	Z.ai	95
GLM-5.3 (max), Claude Code	Z.ai	95
Claude Opus 5.5 (medium), Claude Code	Anthropic	95
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	95
Claude Haiku 5.5 (medium), Claude Code	Anthropic	95
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	95
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	95
Claude Sonnet 5.5 (low), Claude Code	Anthropic	95
Claude Sonnet 5 (high), Claude Code	Anthropic	95
GPT-6.1 Sol (xhigh), Codex	OpenAI	95
GPT-5.6 Terra (high), Codex	OpenAI	95
GPT-5.6 Luna (low), Codex	OpenAI	95
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	89
Gemini 3.8 Flash (low), Terminus-2	Google	89
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	89
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	89
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	89
DeepSeek V4 Flash (high), Codex	DeepSeek	89
Claude Opus 5 (high), Claude Code	Anthropic	89
GPT-6 Luna (medium), Codex	OpenAI	89
GPT-5.6 Sol (high), Codex	OpenAI	89
DeepSeek V4.1 Flash (max), Codex	DeepSeek	88
GLM-5.3-Flash (max), Terminus-2	Z.ai	84
GLM-5.3 (max), Terminus-2	Z.ai	84
Claude Haiku 5.5 (high), Claude Code	Anthropic	84
Gemini 3.1 Pro (preview) (low), Terminus-2	Google	79
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	79
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	79
GPT-6 Luna (low), Codex	OpenAI	79
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	74
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	74
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	63
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	63
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	58
gpt-oss-120b (high), Terminus-2	OpenAI	53
Step 5 Preview (Reasoning), Terminus-2	StepFun	53
gpt-oss-120b (low), Terminus-2	OpenAI	47
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	47
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	47
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	42
gpt-oss-120b (high), Codex	OpenAI	42
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	37
gpt-oss-20b (high), Terminus-2	OpenAI	37
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	37
Gemini 3.8 Flash (high), Terminus-2	Google	32
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	32
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	26
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	16
gpt-oss-20b (high), Codex	OpenAI	11
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	0

Open weights

GeoAnalystBench score

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

How to read this chart

About this evaluation. Spatial workflow and code generation. GeoAnalystBench results and background.

Data table: GeoAnalystBench score

GeoAnalystBench score
Model	Developer	GeoAnalystBench score
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	100
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	100
GLM-5.2 (high), Terminus-2	Z.ai	100
GLM-5.3-Flash (max), Claude Code	Z.ai	100
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	95
GLM-5.2 (max), Terminus-2	Z.ai	95
GLM-5.3-Flash (low), Terminus-2	Z.ai	95
GLM-5.3 (low), Terminus-2	Z.ai	95
GLM-5.3 (max), Claude Code	Z.ai	95
Claude Sonnet 5.5 (low), Claude Code	Anthropic	95
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	89
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	89
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	89
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	89
DeepSeek V4 Flash (high), Codex	DeepSeek	89
DeepSeek V4.1 Flash (max), Codex	DeepSeek	88
GLM-5.3-Flash (max), Terminus-2	Z.ai	84
GLM-5.3 (max), Terminus-2	Z.ai	84
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	79
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	79
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	74
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	63
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	63
gpt-oss-120b (high), Terminus-2	OpenAI	53
Step 5 Preview (Reasoning), Terminus-2	StepFun	53
gpt-oss-120b (low), Terminus-2	OpenAI	47
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	47
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	47
gpt-oss-120b (high), Codex	OpenAI	42
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	37
gpt-oss-20b (high), Terminus-2	OpenAI	37
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	37
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	32
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	26
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	16
gpt-oss-20b (high), Codex	OpenAI	11
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	0

Vision

All

Vision agents score

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

How to read this chart

Vision agents. Agents that look at aerial images and use code to find, count and measure what the question asks. Vision agents results.

Data table: Vision agents score

Vision agents score
Model	Developer	Vision agents score
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	97
GPT-6 Sol (medium), Codex	OpenAI	91
GPT-6.1 Sol (xhigh), Codex	OpenAI	90
Claude Opus 5.5 (high), Claude Code	Anthropic	89
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	89
Claude Opus 5.5 (low), Claude Code	Anthropic	89
Claude Opus 5 (high), Claude Code	Anthropic	86
GPT-6.1 Sol (high), Codex	OpenAI	86
Claude Sonnet 5 (high), Claude Code	Anthropic	81
DeepSeek V4.1 Flash (max), Codex	DeepSeek	80
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	80
GPT-6 Sol (high), Codex	OpenAI	80
GLM-5.3-Flash (max), Claude Code	Z.ai	78
Claude Sonnet 5.5 (high), Claude Code	Anthropic	78
Claude Sonnet 5.5 (low), Claude Code	Anthropic	78
Claude Haiku 5.5 (low), Claude Code	Anthropic	77
Claude Opus 4.7 (high), Claude Code	Anthropic	77
GPT-5.6 Sol (high), Codex	OpenAI	75
Claude Opus 4.8 (high), Claude Code	Anthropic	73
GPT-5.6 Luna (xhigh), Codex	OpenAI	72
Claude Haiku 5.5 (high), Claude Code	Anthropic	70
GPT-6 Luna (xhigh), Codex	OpenAI	70
GPT-5.6 Terra (high), Codex	OpenAI	69
GPT-5.6 Luna (medium), Codex	OpenAI	69
GPT-5.6 Luna (high), Codex	OpenAI	69
GPT-6 Luna (high), Codex	OpenAI	67
GPT-6 Luna (medium), Codex	OpenAI	64
GPT-5.6 Luna (low), Codex	OpenAI	64
GPT-6 Luna (low), Codex	OpenAI	28
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	19

Open weights

Vision agents score

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

How to read this chart

Vision agents. Agents that look at aerial images and use code to find, count and measure what the question asks. Vision agents results.

Data table: Vision agents score

Vision agents score
Model	Developer	Vision agents score
DeepSeek V4.1 Flash (max), Codex	DeepSeek	80
GLM-5.3-Flash (max), Claude Code	Z.ai	78
Claude Sonnet 5.5 (low), Claude Code	Anthropic	78

Knowledge

All

Geospatial knowledge

Share of questions answered correctly, 0 to 100 · Higher is better

How to read this chart

What this metric means. Closed-book multiple-choice questions on GIS and Earth observation. Geospatial knowledge results.

Data table: Geospatial knowledge

Geospatial knowledge
Model	Developer	Geospatial knowledge
Claude Sonnet 4.6 (high)	Anthropic	97.9
Claude Sonnet 5.5 (low)	Anthropic	97.5
Claude Opus 5.5 (xhigh)	Anthropic	94.3
Claude Sonnet 5.5 (xhigh)	Anthropic	94.1
Gemini 3.8 Flash (low)	Google	93.7
Claude Opus 4.8 (high)	Anthropic	93.3
Claude Opus 5 (high)	Anthropic	93.1
Claude Opus 4.7 (high)	Anthropic	92.8
GPT-5.6 Sol (high)	OpenAI	92.6
GPT-6.1 Sol (xhigh)	OpenAI	92.5
GPT-5.6 Terra (high)	OpenAI	92.4
Kimi K2.7 Code (Reasoning) Open weights	Moonshot AI	92.1
Kimi K3 (Reasoning) Open weights	Moonshot AI	92.0
DeepSeek V4 Pro (max) Open weights	DeepSeek	91.9
Claude Haiku 5.5 (xhigh)	Anthropic	91.8
GLM-5.3 (max) Open weights	Z.ai	91.8
Kimi K2.6 (Reasoning) Open weights	Moonshot AI	91.8
GLM-5.3-Flash (max) Open weights	Z.ai	91.8
DeepSeek V4 Flash (max) Open weights	DeepSeek	91.5
Claude Sonnet 5 (high)	Anthropic	91.4
GPT-6 Luna (xhigh)	OpenAI	91.3
Qwen 3.8 27B (xhigh) Open weights	Alibaba (Qwen)	91.1
GPT-6 Sol (high)	OpenAI	91.0
Nemotron 3 Ultra (high) Open weights	NVIDIA	90.7
GPT-5.6 Luna (xhigh)	OpenAI	90.7
Gemini 3.1 Flash-Lite (high)	Google	90.0
gpt-oss-120b (high) Open weights	OpenAI	89.8
Claude Haiku 4.5 (Reasoning)	Anthropic	89.0
GLM-5.2 (max) Open weights	Z.ai	89.0
Nemotron 3.5 Super VL 120B (Reasoning) Open weights	NVIDIA	89.0
Mistral Large 3 (Non-reasoning) Open weights	Mistral AI	88.7
Gemma 4 26B A4B (Reasoning) Open weights	Google	88.7
Nemotron 3 Nano Omni 30B (Reasoning) Open weights	NVIDIA	86.3
GLM-4.7-Flash (Reasoning) Open weights	Z.ai	85.6
gpt-oss-20b (high) Open weights	OpenAI	84.8
Llama 3.3 70B (Non-reasoning) Open weights	Meta	82.5
Mistral Small 3.1 24B (Non-reasoning) Open weights	Mistral AI	81.8

Open weights

Geospatial knowledge

Share of questions answered correctly, 0 to 100 · Higher is better

How to read this chart

What this metric means. Closed-book multiple-choice questions on GIS and Earth observation. Geospatial knowledge results.

Data table: Geospatial knowledge

Geospatial knowledge
Model	Developer	Geospatial knowledge
Claude Sonnet 5.5 (low)	Anthropic	97.5
Kimi K2.7 Code (Reasoning) Open weights	Moonshot AI	92.1
Kimi K3 (Reasoning) Open weights	Moonshot AI	92.0
DeepSeek V4 Pro (max) Open weights	DeepSeek	91.9
GLM-5.3 (max) Open weights	Z.ai	91.8
Kimi K2.6 (Reasoning) Open weights	Moonshot AI	91.8
GLM-5.3-Flash (max) Open weights	Z.ai	91.8
DeepSeek V4 Flash (max) Open weights	DeepSeek	91.5
Qwen 3.8 27B (xhigh) Open weights	Alibaba (Qwen)	91.1
Nemotron 3 Ultra (high) Open weights	NVIDIA	90.7
gpt-oss-120b (high) Open weights	OpenAI	89.8
GLM-5.2 (max) Open weights	Z.ai	89.0
Nemotron 3.5 Super VL 120B (Reasoning) Open weights	NVIDIA	89.0
Mistral Large 3 (Non-reasoning) Open weights	Mistral AI	88.7
Gemma 4 26B A4B (Reasoning) Open weights	Google	88.7
Nemotron 3 Nano Omni 30B (Reasoning) Open weights	NVIDIA	86.3
GLM-4.7-Flash (Reasoning) Open weights	Z.ai	85.6
gpt-oss-20b (high) Open weights	OpenAI	84.8
Llama 3.3 70B (Non-reasoning) Open weights	Meta	82.5
Mistral Small 3.1 24B (Non-reasoning) Open weights	Mistral AI	81.8

Standalone evaluations

A standalone evaluation: its scores are shown here and on model pages, but it is not counted in any index or ranking until more models have run it.

All

OGD4All score

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

How to read this chart

About this evaluation. Questions on a city's open government data. OGD4All results and background.

Data table: OGD4All score

OGD4All score
Model	Developer	OGD4All score
Claude Opus 4.8 (high), Claude Code	Anthropic	97
GLM-5.3-Flash (max), Claude Code	Z.ai	97
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	97
Claude Opus 5.5 (medium), Claude Code	Anthropic	97
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	97
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	97
Claude Haiku 5.5 (high), Claude Code	Anthropic	96
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	96
Claude Sonnet 5.5 (high), Claude Code	Anthropic	96
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	96
Claude Sonnet 5.5 (low), Claude Code	Anthropic	96
Claude Opus 5.5 (high), Claude Code	Anthropic	95
Claude Haiku 5.5 (medium), Claude Code	Anthropic	95
Claude Opus 5.5 (low), Claude Code	Anthropic	95
Claude Opus 5 (high), Claude Code	Anthropic	95
Claude Sonnet 5 (high), Claude Code	Anthropic	95
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	95
Claude Haiku 5.5 (low), Claude Code	Anthropic	95
Claude Opus 4.7 (high), Claude Code	Anthropic	95

Open weights

OGD4All score

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

How to read this chart

About this evaluation. Questions on a city's open government data. OGD4All results and background.

Data table: OGD4All score

OGD4All score
Model	Developer	OGD4All score
GLM-5.3-Flash (max), Claude Code	Z.ai	97
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	97
Claude Sonnet 5.5 (low), Claude Code	Anthropic	96

All

GeoNatureAgent score

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

How to read this chart

About this evaluation. Environmental raster analysis over named regions. GeoNatureAgent results and background.

Data table: GeoNatureAgent score

GeoNatureAgent score
Model	Developer	GeoNatureAgent score
GLM-5.3-Flash (max), Claude Code	Z.ai	100
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	100
Claude Opus 5.5 (medium), Claude Code	Anthropic	100
Claude Opus 5.5 (high), Claude Code	Anthropic	100
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	100
Claude Sonnet 5.5 (high), Claude Code	Anthropic	100
Claude Haiku 5.5 (medium), Claude Code	Anthropic	100
Claude Haiku 5.5 (high), Claude Code	Anthropic	100
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	100
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	100
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	100
Claude Opus 5.5 (low), Claude Code	Anthropic	100
Claude Sonnet 5.5 (low), Claude Code	Anthropic	100
Claude Haiku 5.5 (low), Claude Code	Anthropic	100
Claude Opus 5 (high), Claude Code	Anthropic	100
Claude Sonnet 5 (high), Claude Code	Anthropic	100
Claude Opus 4.8 (high), Claude Code	Anthropic	100
Claude Opus 4.7 (high), Claude Code	Anthropic	100
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	93

Open weights

GeoNatureAgent score

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

How to read this chart

About this evaluation. Environmental raster analysis over named regions. GeoNatureAgent results and background.

Data table: GeoNatureAgent score

GeoNatureAgent score
Model	Developer	GeoNatureAgent score
GLM-5.3-Flash (max), Claude Code	Z.ai	100
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	100
Claude Sonnet 5.5 (low), Claude Code	Anthropic	100

All

MultiGlobeQA score

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

How to read this chart

About this evaluation. Spatial reasoning over facts about places worldwide. MultiGlobeQA results and background.

Data table: MultiGlobeQA score

MultiGlobeQA score
Model	Developer	MultiGlobeQA score
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	80
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	79
Claude Opus 5 (high), Claude Code	Anthropic	79
Claude Opus 5.5 (medium), Claude Code	Anthropic	79
Claude Opus 5.5 (low), Claude Code	Anthropic	79
Claude Opus 5.5 (high), Claude Code	Anthropic	78
Claude Sonnet 5.5 (low), Claude Code	Anthropic	77
Claude Opus 4.7 (high), Claude Code	Anthropic	77
Claude Sonnet 5.5 (high), Claude Code	Anthropic	77
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	76
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	76
Claude Opus 4.8 (high), Claude Code	Anthropic	75
Claude Haiku 5.5 (high), Claude Code	Anthropic	74
GLM-5.3-Flash (max), Claude Code	Z.ai	74
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	74
Claude Sonnet 5 (high), Claude Code	Anthropic	73
Claude Haiku 5.5 (medium), Claude Code	Anthropic	73
Claude Haiku 5.5 (low), Claude Code	Anthropic	72
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	70

Open weights

MultiGlobeQA score

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

How to read this chart

About this evaluation. Spatial reasoning over facts about places worldwide. MultiGlobeQA results and background.

Data table: MultiGlobeQA score

MultiGlobeQA score
Model	Developer	MultiGlobeQA score
Claude Sonnet 5.5 (low), Claude Code	Anthropic	77
GLM-5.3-Flash (max), Claude Code	Z.ai	74
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	74

All

GeoAgentEval score

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

How to read this chart

About this evaluation. Building, auditing and repairing a PostGIS database. GeoAgentEval results and background.

Data table: GeoAgentEval score

GeoAgentEval score
Model	Developer	GeoAgentEval score
Claude Opus 5.5 (medium), Claude Code	Anthropic	100
Claude Opus 5.5 (high), Claude Code	Anthropic	100
Claude Sonnet 5.5 (high), Claude Code	Anthropic	100
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	100
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	100
Claude Opus 5.5 (low), Claude Code	Anthropic	100
Claude Sonnet 5.5 (low), Claude Code	Anthropic	100
GLM-5.3-Flash (max), Claude Code	Z.ai	96
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	96
Claude Opus 5 (high), Claude Code	Anthropic	96
Claude Haiku 5.5 (medium), Claude Code	Anthropic	92
Claude Haiku 5.5 (high), Claude Code	Anthropic	92
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	92
Claude Haiku 5.5 (low), Claude Code	Anthropic	92
Claude Sonnet 5 (high), Claude Code	Anthropic	92
Claude Opus 4.8 (high), Claude Code	Anthropic	92
Claude Opus 4.7 (high), Claude Code	Anthropic	92
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	88
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	88

Open weights

GeoAgentEval score

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

How to read this chart

About this evaluation. Building, auditing and repairing a PostGIS database. GeoAgentEval results and background.

Data table: GeoAgentEval score

GeoAgentEval score
Model	Developer	GeoAgentEval score
Claude Sonnet 5.5 (low), Claude Code	Anthropic	100
GLM-5.3-Flash (max), Claude Code	Z.ai	96
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	88

Cost

All

Cost per task

Average cost per agent task (USD) · Lower is better

How to read this chart

What cost is measuring. What one attempt at a task would cost if paid token by token at the provider's published API prices, averaged over attempts: ordinary input tokens, input tokens served from the provider's cache at the lower cached price, tokens written to the cache where the provider charges for that, and output tokens (thinking included). Cost assumes each harness uses the model's pay-per-token API, priced at public list rates; sandboxes and our own engineering time are not included. Failed attempts count too.

How to read this chart. Lower is better. Costs are pay-per-token list prices, not any discount or credit a customer might have.

Data table: Cost per task

Cost per task
Model	Developer	Cost per task
GPT-6 Luna (medium), Codex	OpenAI	$0.0026
GPT-6 Luna (low), Codex	OpenAI	$0.0032
GPT-6 Luna (high), Codex	OpenAI	$0.0040
gpt-oss-120b (low), Terminus-2	OpenAI	$0.0046
Claude Haiku 5.5 (low), Claude Code	Anthropic	$0.0053
GPT-5.6 Luna (low), Codex	OpenAI	$0.0058
Claude Haiku 5.5 (medium), Claude Code	Anthropic	$0.0063
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	$0.0066
GPT-6 Luna (xhigh), Codex	OpenAI	$0.0070
Claude Haiku 5.5 (high), Claude Code	Anthropic	$0.0085
GPT-5.6 Luna (medium), Codex	OpenAI	$0.0094
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	$0.0096
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	$0.012
GPT-5.6 Luna (high), Codex	OpenAI	$0.016
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	$0.017
gpt-oss-20b (high), Terminus-2	OpenAI	$0.017
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	$0.017
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	$0.018
GPT-5.6 Luna (xhigh), Codex	OpenAI	$0.021
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	$0.021
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	$0.023
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	$0.026
gpt-oss-120b (high), Terminus-2	OpenAI	$0.026
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	$0.031
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	$0.037
Claude Sonnet 5.5 (low), Claude Code	Anthropic	$0.043
GPT-6.1 Sol (medium), Codex	OpenAI	$0.046
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	$0.046
GLM-5.3-Flash (low), Terminus-2	Z.ai	$0.048
Claude Sonnet 5.5 (high), Claude Code	Anthropic	$0.055
GPT-6.1 Sol (high), Codex	OpenAI	$0.059
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	$0.062
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	$0.065
DeepSeek V4.1 Flash (max), Codex	DeepSeek	$0.067
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	$0.069
Step 5 Preview (Reasoning), Terminus-2	StepFun	$0.071
Claude Opus 5.5 (low), Claude Code	Anthropic	$0.081
GPT-6 Sol (high), Codex	OpenAI	$0.082
GPT-6.1 Sol (xhigh), Codex	OpenAI	$0.085
DeepSeek V4 Flash (high), Codex	DeepSeek	$0.085
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	$0.089
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	$0.092
GLM-5.3-Flash (max), Claude Code	Z.ai	$0.093
gpt-oss-20b (high), Codex	OpenAI	$0.094
GLM-5.3-Flash (max), Terminus-2	Z.ai	$0.094
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	$0.099
GPT-5.6 Terra (high), Codex	OpenAI	$0.10
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	$0.10
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	$0.11
Claude Opus 5.5 (medium), Claude Code	Anthropic	$0.11
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	$0.11
Gemini 3.8 Flash (low), Terminus-2	Google	$0.11
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	$0.13
Claude Opus 5.5 (high), Claude Code	Anthropic	$0.13
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	$0.18
Claude Sonnet 5 (high), Claude Code	Anthropic	$0.20
GPT-5.6 Sol (high), Codex	OpenAI	$0.21
GLM-5.2 (high), Terminus-2	Z.ai	$0.22
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	$0.22
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	$0.23
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	$0.25
gpt-oss-120b (high), Codex	OpenAI	$0.26
Claude Opus 4.8 (high), Claude Code	Anthropic	$0.26
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	$0.27
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	$0.27
GLM-5.3 (low), Terminus-2	Z.ai	$0.27
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	$0.30
Claude Opus 5 (high), Claude Code	Anthropic	$0.32
GLM-5.2 (max), Terminus-2	Z.ai	$0.32
Claude Opus 4.7 (high), Claude Code	Anthropic	$0.34
GLM-5.2 (max), Claude Code	Z.ai	$0.34
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	$0.35
GLM-5.3 (max), Terminus-2	Z.ai	$0.38
GLM-5.3 (max), Claude Code	Z.ai	$0.48

Open weights

Cost per task

Average cost per agent task (USD) · Lower is better

How to read this chart

How to read this chart. Lower is better. Costs are pay-per-token list prices, not any discount or credit a customer might have.

Data table: Cost per task

Cost per task
Model	Developer	Cost per task
gpt-oss-120b (low), Terminus-2	OpenAI	$0.0046
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	$0.0096
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	$0.012
gpt-oss-20b (high), Terminus-2	OpenAI	$0.017
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	$0.017
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	$0.021
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	$0.023
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	$0.026
gpt-oss-120b (high), Terminus-2	OpenAI	$0.026
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	$0.037
Claude Sonnet 5.5 (low), Claude Code	Anthropic	$0.043
GLM-5.3-Flash (low), Terminus-2	Z.ai	$0.048
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	$0.062
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	$0.065
DeepSeek V4.1 Flash (max), Codex	DeepSeek	$0.067
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	$0.069
Step 5 Preview (Reasoning), Terminus-2	StepFun	$0.071
DeepSeek V4 Flash (high), Codex	DeepSeek	$0.085
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	$0.089
GLM-5.3-Flash (max), Claude Code	Z.ai	$0.093
gpt-oss-20b (high), Codex	OpenAI	$0.094
GLM-5.3-Flash (max), Terminus-2	Z.ai	$0.094
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	$0.099
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	$0.10
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	$0.11
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	$0.13
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	$0.18
GLM-5.2 (high), Terminus-2	Z.ai	$0.22
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	$0.22
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	$0.23
gpt-oss-120b (high), Codex	OpenAI	$0.26
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	$0.27
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	$0.27
GLM-5.3 (low), Terminus-2	Z.ai	$0.27
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	$0.30
GLM-5.2 (max), Terminus-2	Z.ai	$0.32
GLM-5.2 (max), Claude Code	Z.ai	$0.34
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	$0.35
GLM-5.3 (max), Terminus-2	Z.ai	$0.38
GLM-5.3 (max), Claude Code	Z.ai	$0.48

All

Geospatial Agent Index vs. cost per task

Geospatial Agent Index, Text only; cost per task (USD) · Higher and further left is better

Most attractive quadrant
Pareto line

How to read this chart

Why a log scale. Costs and times differ by a factor of ten or more between models, so equal steps on the axis are equal ratios.

Not yet ranked. Entries with too few tasks run so far are not plotted; the table lists them, and every entry is on the models page.

How to read this chart. Each point is one entry. Further left means a lower average cost per task (log scale); higher means a higher Geospatial Agent Index. Both axes fit the models shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot, split at the middle of each axis: a reading aid that moves with the axes when you change the models shown, not a fixed bar. The dotted Pareto line joins the shown models that no cheaper shown model beats on score.

Data table: Geospatial Agent Index vs. cost per task

Geospatial Agent Index vs. cost per task
Model	Geospatial Agent Index	Cost per task (USD)	On the Pareto line
DeepSeek V4 Pro (max), Terminus-2	71	$0.35	No
Gemini 3.1 Flash-Lite (high), Terminus-2	56	$0.031	No
Gemini 3.1 Pro (preview) (high), Terminus-2	66	$0.22	Not plotted: not yet ranked
Gemini 3.8 Flash (low), Terminus-2	71	$0.11	No
Gemma 4 26B A4B (Reasoning), Terminus-2	56	$0.021	No
gpt-oss-120b (high), Terminus-2	46	$0.026	No
gpt-oss-20b (high), Terminus-2	33	$0.017	No
Kimi K3 (Reasoning), Terminus-2	72	$0.23	No
Llama 3.3 70B (Non-reasoning), Terminus-2	8	$0.023	No
MiniMax M3.1 Flash (max), Terminus-2	51	$0.017	No
Mistral Large 3 (Non-reasoning), Terminus-2	33	$0.026	No
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	20	$0.13	No
Nemotron 3 Ultra (high), Terminus-2	68	$0.069	No
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	11	$0.012	No
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	52	$0.0096	Yes
Step 5 Preview (Reasoning), Terminus-2	34	$0.071	No
GLM-5.3-Flash (max), Claude Code	72	$0.093	No
GLM-5.3 (max), Claude Code	73	$0.48	No
DeepSeek V4.1 Flash (max), Codex	71	$0.067	No
Kimi K2.7 Code (Reasoning), Kimi Code	72	$0.18	No
Qwen 3.8 27B (xhigh), Qwen Code	68	$0.22	No
Claude Opus 5.5 (xhigh), Claude Code	85	$0.25	Yes
Claude Sonnet 5.5 (xhigh), Claude Code	79	$0.11	No
Claude Haiku 5.5 (xhigh), Claude Code	76	$0.017	Yes
Claude Sonnet 5.5 (low), Claude Code	69	$0.043	No
GPT-6.1 Sol (xhigh), Codex	82	$0.085	Yes
GPT-5.6 Terra (high), Codex	69	$0.10	No
GPT-5.6 Luna (xhigh), Codex	73	$0.021	No

Open weights

Geospatial Agent Index vs. cost per task

Geospatial Agent Index, Text only; cost per task (USD) · Higher and further left is better

Most attractive quadrant
Pareto line

How to read this chart

Why a log scale. Costs and times differ by a factor of ten or more between models, so equal steps on the axis are equal ratios.

Data table: Geospatial Agent Index vs. cost per task

Geospatial Agent Index vs. cost per task
Model	Geospatial Agent Index	Cost per task (USD)	On the Pareto line
DeepSeek V4 Pro (max), Terminus-2	71	$0.35	No
Gemma 4 26B A4B (Reasoning), Terminus-2	56	$0.021	Yes
gpt-oss-120b (high), Terminus-2	46	$0.026	No
gpt-oss-20b (high), Terminus-2	33	$0.017	No
Kimi K3 (Reasoning), Terminus-2	72	$0.23	Yes
Llama 3.3 70B (Non-reasoning), Terminus-2	8	$0.023	No
MiniMax M3.1 Flash (max), Terminus-2	51	$0.017	No
Mistral Large 3 (Non-reasoning), Terminus-2	33	$0.026	No
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	20	$0.13	No
Nemotron 3 Ultra (high), Terminus-2	68	$0.069	No
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	11	$0.012	No
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	52	$0.0096	Yes
Step 5 Preview (Reasoning), Terminus-2	34	$0.071	No
GLM-5.3-Flash (max), Claude Code	72	$0.093	Yes
GLM-5.3 (max), Claude Code	73	$0.48	Yes
DeepSeek V4.1 Flash (max), Codex	71	$0.067	Yes
Kimi K2.7 Code (Reasoning), Kimi Code	72	$0.18	Yes
Qwen 3.8 27B (xhigh), Qwen Code	68	$0.22	No
Claude Sonnet 5.5 (low), Claude Code	69	$0.043	Yes

Time

All

Time per task

Average minutes per agent task, at 100 output tokens/s · Lower is better

How to read this chart

What this metric means. Time as if every model wrote 100 output tokens per second: output tokens divided by 100, plus the time the agent's commands took. Serving speed differs between providers and is not ranked.

Data table: Time per task

Time per task
Model	Developer	Time per task (at 100 output tokens/s)
Claude Opus 5.5 (low), Claude Code	Anthropic	0.4 min
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	0.4 min
GPT-6.1 Sol (medium), Codex	OpenAI	0.4 min
GPT-6 Luna (medium), Codex	OpenAI	0.4 min
Claude Sonnet 5.5 (low), Claude Code	Anthropic	0.4 min
GPT-6 Luna (low), Codex	OpenAI	0.5 min
gpt-oss-120b (low), Terminus-2	OpenAI	0.5 min
GPT-6.1 Sol (high), Codex	OpenAI	0.5 min
Claude Opus 5.5 (medium), Claude Code	Anthropic	0.5 min
GPT-5.6 Luna (low), Codex	OpenAI	0.5 min
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	0.6 min
Claude Sonnet 5.5 (high), Claude Code	Anthropic	0.6 min
GPT-6 Luna (high), Codex	OpenAI	0.7 min
Claude Opus 5.5 (high), Claude Code	Anthropic	0.7 min
GPT-6 Sol (high), Codex	OpenAI	0.8 min
GPT-5.6 Luna (medium), Codex	OpenAI	0.8 min
GPT-6.1 Sol (xhigh), Codex	OpenAI	0.8 min
GPT-5.6 Terra (high), Codex	OpenAI	0.8 min
Claude Opus 4.8 (high), Claude Code	Anthropic	0.9 min
Claude Haiku 5.5 (low), Claude Code	Anthropic	0.9 min
Claude Opus 4.7 (high), Claude Code	Anthropic	0.9 min
Claude Haiku 5.5 (medium), Claude Code	Anthropic	1.0 min
GPT-5.6 Sol (high), Codex	OpenAI	1.0 min
Claude Opus 5 (high), Claude Code	Anthropic	1.1 min
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	1.1 min
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	1.2 min
Claude Haiku 5.5 (high), Claude Code	Anthropic	1.4 min
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	1.4 min
GPT-6 Luna (xhigh), Codex	OpenAI	1.4 min
GPT-5.6 Luna (high), Codex	OpenAI	1.4 min
Claude Sonnet 5 (high), Claude Code	Anthropic	1.5 min
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	1.5 min
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	1.6 min
GPT-5.6 Luna (xhigh), Codex	OpenAI	1.7 min
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	1.7 min
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	1.9 min
Gemini 3.8 Flash (low), Terminus-2	Google	1.9 min
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2.1 min
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	2.4 min
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	2.5 min
GLM-5.3 (low), Terminus-2	Z.ai	2.7 min
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	2.7 min
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	3.0 min
gpt-oss-120b (high), Terminus-2	OpenAI	3.0 min
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	3.1 min
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	3.5 min
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	3.8 min
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	3.8 min
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	4.0 min
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	4.0 min
GLM-5.3 (max), Terminus-2	Z.ai	4.2 min
GLM-5.3-Flash (low), Terminus-2	Z.ai	4.2 min
GLM-5.3 (max), Claude Code	Z.ai	4.3 min
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	4.7 min
GLM-5.2 (high), Terminus-2	Z.ai	4.7 min
gpt-oss-20b (high), Codex	OpenAI	4.8 min
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	4.9 min
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	4.9 min
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	4.9 min
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	5.4 min
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	5.6 min
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	5.7 min
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	5.9 min
gpt-oss-120b (high), Codex	OpenAI	6.0 min
GLM-5.2 (max), Terminus-2	Z.ai	6.1 min
GLM-5.2 (max), Claude Code	Z.ai	6.3 min
DeepSeek V4 Flash (high), Codex	DeepSeek	6.5 min
GLM-5.3-Flash (max), Claude Code	Z.ai	6.5 min
gpt-oss-20b (high), Terminus-2	OpenAI	7.2 min
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	7.3 min
Step 5 Preview (Reasoning), Terminus-2	StepFun	7.9 min
GLM-5.3-Flash (max), Terminus-2	Z.ai	8.5 min
DeepSeek V4.1 Flash (max), Codex	DeepSeek	8.6 min
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	8.6 min

Open weights

Time per task

Average minutes per agent task, at 100 output tokens/s · Lower is better

How to read this chart

Data table: Time per task

Time per task
Model	Developer	Time per task (at 100 output tokens/s)
Claude Sonnet 5.5 (low), Claude Code	Anthropic	0.4 min
gpt-oss-120b (low), Terminus-2	OpenAI	0.5 min
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	1.1 min
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	1.5 min
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	1.9 min
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2.1 min
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	2.4 min
GLM-5.3 (low), Terminus-2	Z.ai	2.7 min
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	2.7 min
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	3.0 min
gpt-oss-120b (high), Terminus-2	OpenAI	3.0 min
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	3.5 min
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	3.8 min
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	3.8 min
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	4.0 min
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	4.0 min
GLM-5.3 (max), Terminus-2	Z.ai	4.2 min
GLM-5.3-Flash (low), Terminus-2	Z.ai	4.2 min
GLM-5.3 (max), Claude Code	Z.ai	4.3 min
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	4.7 min
GLM-5.2 (high), Terminus-2	Z.ai	4.7 min
gpt-oss-20b (high), Codex	OpenAI	4.8 min
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	4.9 min
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	4.9 min
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	4.9 min
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	5.4 min
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	5.6 min
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	5.7 min
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	5.9 min
gpt-oss-120b (high), Codex	OpenAI	6.0 min
GLM-5.2 (max), Terminus-2	Z.ai	6.1 min
GLM-5.2 (max), Claude Code	Z.ai	6.3 min
DeepSeek V4 Flash (high), Codex	DeepSeek	6.5 min
GLM-5.3-Flash (max), Claude Code	Z.ai	6.5 min
gpt-oss-20b (high), Terminus-2	OpenAI	7.2 min
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	7.3 min
Step 5 Preview (Reasoning), Terminus-2	StepFun	7.9 min
GLM-5.3-Flash (max), Terminus-2	Z.ai	8.5 min
DeepSeek V4.1 Flash (max), Codex	DeepSeek	8.6 min
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	8.6 min

All

Geospatial Agent Index vs. time per task

Geospatial Agent Index, Text only; time per task (minutes) · Higher and further left is better

Most attractive quadrant
Pareto line

How to read this chart

Why a log scale. Costs and times differ by a factor of ten or more between models, so equal steps on the axis are equal ratios.

Not yet ranked. Entries with too few tasks run so far are not plotted; the table lists them, and every entry is on the models page.

How to read this chart. Each point is one entry. Further left means a lower average time per task (log scale); higher means a higher Geospatial Agent Index. Both axes fit the models shown, so the score axis does not start at zero. The shaded most attractive quadrant is the top-left quarter of the plot, split at the middle of each axis: a reading aid that moves with the axes when you change the models shown, not a fixed bar. The dotted Pareto line joins the shown models that no faster shown model beats on score.

Data table: Geospatial Agent Index vs. time per task

Geospatial Agent Index vs. time per task
Model	Geospatial Agent Index	Time per task (minutes)	On the Pareto line
DeepSeek V4 Pro (max), Terminus-2	71	4.9 min	No
Gemini 3.1 Flash-Lite (high), Terminus-2	56	3.1 min	No
Gemini 3.1 Pro (preview) (high), Terminus-2	66	2.5 min	Not plotted: not yet ranked
Gemini 3.8 Flash (low), Terminus-2	71	1.9 min	No
Gemma 4 26B A4B (Reasoning), Terminus-2	56	5.6 min	No
gpt-oss-120b (high), Terminus-2	46	3.0 min	No
gpt-oss-20b (high), Terminus-2	33	7.2 min	No
Kimi K3 (Reasoning), Terminus-2	72	2.7 min	No
Llama 3.3 70B (Non-reasoning), Terminus-2	8	2.1 min	No
MiniMax M3.1 Flash (max), Terminus-2	51	8.6 min	No
Mistral Large 3 (Non-reasoning), Terminus-2	33	1.1 min	No
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	20	4.9 min	No
Nemotron 3 Ultra (high), Terminus-2	68	2.4 min	No
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	11	4.9 min	No
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	52	3.8 min	No
Step 5 Preview (Reasoning), Terminus-2	34	7.9 min	No
GLM-5.3-Flash (max), Claude Code	72	6.5 min	No
GLM-5.3 (max), Claude Code	73	4.3 min	No
DeepSeek V4.1 Flash (max), Codex	71	8.6 min	No
Kimi K2.7 Code (Reasoning), Kimi Code	72	1.9 min	No
Qwen 3.8 27B (xhigh), Qwen Code	68	4.7 min	No
Claude Opus 5.5 (xhigh), Claude Code	85	1.4 min	Yes
Claude Sonnet 5.5 (xhigh), Claude Code	79	1.2 min	No
Claude Haiku 5.5 (xhigh), Claude Code	76	2.5 min	No
Claude Sonnet 5.5 (low), Claude Code	69	0.4 min	Yes
GPT-6.1 Sol (xhigh), Codex	82	0.8 min	Yes
GPT-5.6 Terra (high), Codex	69	0.8 min	No
GPT-5.6 Luna (xhigh), Codex	73	1.7 min	No

Open weights

Geospatial Agent Index vs. time per task

Geospatial Agent Index, Text only; time per task (minutes) · Higher and further left is better

Most attractive quadrant
Pareto line

How to read this chart

Why a log scale. Costs and times differ by a factor of ten or more between models, so equal steps on the axis are equal ratios.

Data table: Geospatial Agent Index vs. time per task

Geospatial Agent Index vs. time per task
Model	Geospatial Agent Index	Time per task (minutes)	On the Pareto line
DeepSeek V4 Pro (max), Terminus-2	71	4.9 min	No
Gemma 4 26B A4B (Reasoning), Terminus-2	56	5.6 min	No
gpt-oss-120b (high), Terminus-2	46	3.0 min	No
gpt-oss-20b (high), Terminus-2	33	7.2 min	No
Kimi K3 (Reasoning), Terminus-2	72	2.7 min	Yes
Llama 3.3 70B (Non-reasoning), Terminus-2	8	2.1 min	No
MiniMax M3.1 Flash (max), Terminus-2	51	8.6 min	No
Mistral Large 3 (Non-reasoning), Terminus-2	33	1.1 min	No
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	20	4.9 min	No
Nemotron 3 Ultra (high), Terminus-2	68	2.4 min	No
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	11	4.9 min	No
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	52	3.8 min	No
Step 5 Preview (Reasoning), Terminus-2	34	7.9 min	No
GLM-5.3-Flash (max), Claude Code	72	6.5 min	No
GLM-5.3 (max), Claude Code	73	4.3 min	Yes
DeepSeek V4.1 Flash (max), Codex	71	8.6 min	No
Kimi K2.7 Code (Reasoning), Kimi Code	72	1.9 min	Yes
Qwen 3.8 27B (xhigh), Qwen Code	68	4.7 min	No
Claude Sonnet 5.5 (low), Claude Code	69	0.4 min	Yes

Tokens

All

Output tokens per task

Average output tokens per agent task, reasoning included · Lower is better

How to read this chart

Output tokens per task. The average number of tokens the model wrote per attempt, its reasoning included, as reported by the provider.

Compare with care. Models use different tokenisers, so the same text is a different number of tokens for different models. Use token counts to see how much work each model did, and cost to compare spending.

Data table: Output tokens per task

Output tokens per task
Model	Developer	Output tokens per task
Claude Opus 5.5 (low), Claude Code	Anthropic	1k
Claude Sonnet 5.5 (low), Claude Code	Anthropic	1k
GPT-6 Luna (medium), Codex	OpenAI	1k
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	1k
GPT-6.1 Sol (medium), Codex	OpenAI	1k
GPT-6 Luna (low), Codex	OpenAI	1k
GPT-5.6 Luna (low), Codex	OpenAI	2k
gpt-oss-120b (low), Terminus-2	OpenAI	2k
Claude Opus 5.5 (medium), Claude Code	Anthropic	2k
Claude Sonnet 5.5 (high), Claude Code	Anthropic	2k
GPT-6.1 Sol (high), Codex	OpenAI	2k
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	2k
Claude Opus 5.5 (high), Claude Code	Anthropic	2k
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2k
GPT-6 Luna (high), Codex	OpenAI	3k
GPT-5.6 Luna (medium), Codex	OpenAI	3k
GPT-6 Sol (high), Codex	OpenAI	3k
Claude Haiku 5.5 (low), Claude Code	Anthropic	3k
GPT-5.6 Terra (high), Codex	OpenAI	3k
GPT-6.1 Sol (xhigh), Codex	OpenAI	3k
Claude Haiku 5.5 (medium), Claude Code	Anthropic	4k
GPT-5.6 Sol (high), Codex	OpenAI	4k
Claude Opus 4.8 (high), Claude Code	Anthropic	4k
Claude Opus 4.7 (high), Claude Code	Anthropic	4k
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	4k
Claude Opus 5 (high), Claude Code	Anthropic	4k
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	4k
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	5k
GPT-5.6 Luna (high), Codex	OpenAI	5k
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	6k
Claude Haiku 5.5 (high), Claude Code	Anthropic	6k
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	6k
GPT-6 Luna (xhigh), Codex	OpenAI	6k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	6k
Claude Sonnet 5 (high), Claude Code	Anthropic	6k
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	7k
GPT-5.6 Luna (xhigh), Codex	OpenAI	7k
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	7k
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	8k
Gemini 3.8 Flash (low), Terminus-2	Google	8k
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	10k
GLM-5.3 (low), Terminus-2	Z.ai	10k
Step 5 Preview (Reasoning), Terminus-2	StepFun	10k
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	11k
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	12k
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	12k
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	12k
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	14k
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	15k
gpt-oss-120b (high), Terminus-2	OpenAI	15k
GLM-5.3-Flash (low), Terminus-2	Z.ai	15k
GLM-5.3 (max), Claude Code	Z.ai	15k
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	16k
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	18k
DeepSeek V4 Flash (high), Codex	DeepSeek	19k
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	19k
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	19k
GLM-5.3 (max), Terminus-2	Z.ai	19k
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	20k
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	21k
GLM-5.2 (high), Terminus-2	Z.ai	22k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	22k
gpt-oss-20b (high), Codex	OpenAI	23k
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	24k
gpt-oss-120b (high), Codex	OpenAI	27k
GLM-5.2 (max), Claude Code	Z.ai	28k
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	28k
GLM-5.3-Flash (max), Claude Code	Z.ai	29k
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	29k
GLM-5.3-Flash (max), Terminus-2	Z.ai	30k
GLM-5.2 (max), Terminus-2	Z.ai	30k
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	31k
DeepSeek V4.1 Flash (max), Codex	DeepSeek	38k
gpt-oss-20b (high), Terminus-2	OpenAI	40k

Open weights

Output tokens per task

Average output tokens per agent task, reasoning included · Lower is better

How to read this chart

Output tokens per task. The average number of tokens the model wrote per attempt, its reasoning included, as reported by the provider.

Data table: Output tokens per task

Output tokens per task
Model	Developer	Output tokens per task
Claude Sonnet 5.5 (low), Claude Code	Anthropic	1k
gpt-oss-120b (low), Terminus-2	OpenAI	2k
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	2k
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	4k
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	6k
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	6k
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	6k
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	7k
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	10k
GLM-5.3 (low), Terminus-2	Z.ai	10k
Step 5 Preview (Reasoning), Terminus-2	StepFun	10k
Kimi K2.7 Code (Reasoning), Kimi Code	Moonshot AI	11k
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	12k
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	12k
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	14k
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	15k
gpt-oss-120b (high), Terminus-2	OpenAI	15k
GLM-5.3-Flash (low), Terminus-2	Z.ai	15k
GLM-5.3 (max), Claude Code	Z.ai	15k
Kimi K2.6 (Reasoning), Kimi Code	Moonshot AI	18k
DeepSeek V4 Flash (high), Codex	DeepSeek	19k
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	19k
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	19k
GLM-5.3 (max), Terminus-2	Z.ai	19k
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	20k
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	21k
GLM-5.2 (high), Terminus-2	Z.ai	22k
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	22k
gpt-oss-20b (high), Codex	OpenAI	23k
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	24k
gpt-oss-120b (high), Codex	OpenAI	27k
GLM-5.2 (max), Claude Code	Z.ai	28k
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	28k
GLM-5.3-Flash (max), Claude Code	Z.ai	29k
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	29k
GLM-5.3-Flash (max), Terminus-2	Z.ai	30k
GLM-5.2 (max), Terminus-2	Z.ai	30k
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	31k
DeepSeek V4.1 Flash (max), Codex	DeepSeek	38k
gpt-oss-20b (high), Terminus-2	OpenAI	40k

All

Turns per task

Average agent turns per task · Lower is better

How to read this chart

What this metric means. A turn is one model reply plus the commands it ran.

Data table: Turns per task

Turns per task
Model	Developer	Turns per task
gpt-oss-120b (low), Terminus-2	OpenAI	2.9
Claude Opus 5.5 (low), Claude Code	Anthropic	3.5
Claude Sonnet 5.5 (low), Claude Code	Anthropic	3.9
Claude Sonnet 5.5 (medium), Claude Code	Anthropic	4.2
Claude Opus 5.5 (medium), Claude Code	Anthropic	4.8
Claude Sonnet 5.5 (high), Claude Code	Anthropic	4.8
Gemini 3.1 Flash-Lite (minimal), Terminus-2	Google	4.9
gpt-oss-20b (high), Terminus-2	OpenAI	5.2
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	5.2
Claude Opus 5.5 (high), Claude Code	Anthropic	5.3
Claude Haiku 5.5 (low), Claude Code	Anthropic	5.5
GPT-6.1 Sol (medium), Codex	OpenAI	5.8
gpt-oss-120b (high), Terminus-2	OpenAI	5.8
Claude Opus 4.8 (high), Claude Code	Anthropic	6.0
Claude Haiku 5.5 (medium), Claude Code	Anthropic	6.1
GPT-6 Luna (medium), Codex	OpenAI	6.4
GPT-6.1 Sol (high), Codex	OpenAI	6.6
Claude Sonnet 5.5 (xhigh), Claude Code	Anthropic	6.6
Claude Opus 5.5 (xhigh), Claude Code	Anthropic	6.9
GPT-5.6 Luna (low), Codex	OpenAI	7.0
Claude Opus 4.7 (high), Claude Code	Anthropic	7.0
Claude Haiku 5.5 (high), Claude Code	Anthropic	7.1
Claude Opus 5 (high), Claude Code	Anthropic	7.2
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	7.2
GPT-6.1 Sol (xhigh), Codex	OpenAI	7.4
GPT-6 Luna (high), Codex	OpenAI	7.4
GPT-6 Luna (low), Codex	OpenAI	7.7
GPT-6 Sol (high), Codex	OpenAI	8.7
Gemini 3.1 Flash-Lite (low), Terminus-2	Google	8.8
GPT-6 Luna (xhigh), Codex	OpenAI	8.8
GPT-5.6 Luna (medium), Codex	OpenAI	9.4
Claude Sonnet 5 (high), Claude Code	Anthropic	9.5
GPT-5.6 Sol (high), Codex	OpenAI	9.5
Claude Haiku 5.5 (xhigh), Claude Code	Anthropic	9.6
GPT-5.6 Terra (high), Codex	OpenAI	9.6
Gemini 3.1 Flash-Lite (high), Terminus-2	Google	10.1
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	10.1
GLM-5.3 (max), Terminus-2	Z.ai	10.4
GLM-5.3 (low), Terminus-2	Z.ai	10.4
Claude Haiku 4.5 (Reasoning), Claude Code	Anthropic	10.6
Step 5 Preview (Reasoning), Terminus-2	StepFun	11.1
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	12.1
GPT-5.6 Luna (high), Codex	OpenAI	12.3
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	12.8
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	12.8
GLM-5.2 (high), Terminus-2	Z.ai	12.9
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	13.3
GLM-5.3 (max), Claude Code	Z.ai	13.4
GLM-5.2 (max), Terminus-2	Z.ai	13.6
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	13.9
GPT-5.6 Luna (xhigh), Codex	OpenAI	13.9
GLM-5.3-Flash (low), Terminus-2	Z.ai	14.0
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	14.2
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	14.4
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	14.4
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	14.6
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	14.8
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	14.8
GLM-5.2 (max), Claude Code	Z.ai	15.4
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	15.6
GLM-5.3-Flash (max), Terminus-2	Z.ai	15.8
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	17.3
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	17.6
GLM-5.3-Flash (max), Claude Code	Z.ai	18.9
Gemini 3.8 Flash (low), Terminus-2	Google	19.0
DeepSeek V4.1 Flash (max), Codex	DeepSeek	23.4
gpt-oss-20b (high), Codex	OpenAI	24.2
DeepSeek V4 Flash (high), Codex	DeepSeek	25.4
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	26.4
gpt-oss-120b (high), Codex	OpenAI	29.8
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	30.7
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	37.8

Open weights

Turns per task

Average agent turns per task · Lower is better

How to read this chart

What this metric means. A turn is one model reply plus the commands it ran.

Data table: Turns per task

Turns per task
Model	Developer	Turns per task
gpt-oss-120b (low), Terminus-2	OpenAI	2.9
Claude Sonnet 5.5 (low), Claude Code	Anthropic	3.9
gpt-oss-20b (high), Terminus-2	OpenAI	5.2
Mistral Large 3 (Non-reasoning), Terminus-2	Mistral AI	5.2
gpt-oss-120b (high), Terminus-2	OpenAI	5.8
Kimi K3 (Reasoning), Terminus-2	Moonshot AI	7.2
Nemotron 3.5 Super VL 120B (Reasoning), Terminus-2	NVIDIA	10.1
GLM-5.3 (max), Terminus-2	Z.ai	10.4
GLM-5.3 (low), Terminus-2	Z.ai	10.4
Step 5 Preview (Reasoning), Terminus-2	StepFun	11.1
Nemotron 3 Ultra (high), Terminus-2	NVIDIA	12.1
GLM-5.2 (Non-reasoning), Terminus-2	Z.ai	12.8
Kimi K2.6 (Reasoning), Terminus-2	Moonshot AI	12.8
GLM-5.2 (high), Terminus-2	Z.ai	12.9
MiniMax M3.1 Flash (max), Terminus-2	MiniMax	13.3
GLM-5.3 (max), Claude Code	Z.ai	13.4
GLM-5.2 (max), Terminus-2	Z.ai	13.6
DeepSeek V4 Pro (high), Terminus-2	DeepSeek	13.9
GLM-5.3-Flash (low), Terminus-2	Z.ai	14.0
Kimi K2.7 Code (Reasoning), Terminus-2	Moonshot AI	14.2
Llama 3.3 70B (Non-reasoning), Terminus-2	Meta	14.4
DeepSeek V4 Flash (max), Terminus-2	DeepSeek	14.4
Gemma 4 26B A4B (Reasoning), Terminus-2	Google	14.6
DeepSeek V4 Pro (max), Terminus-2	DeepSeek	14.8
Qwen 3.8 27B (xhigh), Terminus-2	Alibaba (Qwen)	14.8
GLM-5.2 (max), Claude Code	Z.ai	15.4
DeepSeek V4 Flash (high), Terminus-2	DeepSeek	15.6
GLM-5.3-Flash (max), Terminus-2	Z.ai	15.8
Nemotron 3.5 Lightning 30B (Reasoning), Terminus-2	NVIDIA	17.3
Qwen 3.8 27B (xhigh), Qwen Code	Alibaba (Qwen)	17.6
GLM-5.3-Flash (max), Claude Code	Z.ai	18.9
DeepSeek V4.1 Flash (max), Codex	DeepSeek	23.4
gpt-oss-20b (high), Codex	OpenAI	24.2
DeepSeek V4 Flash (high), Codex	DeepSeek	25.4
GLM-4.7-Flash (Reasoning), Terminus-2	Z.ai	26.4
gpt-oss-120b (high), Codex	OpenAI	29.8
Nemotron 3 Nano Omni 30B (Reasoning), Terminus-2	NVIDIA	30.7
GLM-4.7-Flash (Reasoning), Claude Code	Z.ai	37.8