Terminus-2 – Kimi K3 (always on) vs. Codex – GPT-6 Luna (high): geospatial agent comparison

Comparison between Terminus-2 – Kimi K3 (always on) and Codex – GPT-6 Luna (high) on the Geospatial Agent Index, each benchmark, cost, time and token use, from the same tasks, agent and limits. Methodology.

Side by side

Terminus-2 – Kimi K3 (always on) vs. Codex – GPT-6 Luna (high)
Metric	Terminus-2 – Kimi K3 (always on)	Codex – GPT-6 Luna (high)
Geospatial Agent Index	72*	71* (partial: attempt 1 on 444 of 488 tasks)
Earth-Bench V2 score	72	66
GeoBenchX score	64	71
GS-QA2 score	80	77
GeoAgentBench score (guided GIS workflow; outside the index)	94	96
GeoAnalystBench score (guided GIS workflow; outside the index)	89	100
Raster and remote sensing score	80	73
Vector and overlay score	76	77
Climate and time series score	71	69
Spatial statistics and interpolation score	54	33
Mapping and cartography score	81	71
Recognising infeasible requests score	50	78
Cost per task	$0.23	$0.0039
Time per task	3.3 min	1.7 min
Tokens per attempt	53k	116k
Turns per attempt	7.2	7.3
Attempts decided	488	430
Input price per 1M tokens	$3.00	$0.10
Output price per 1M tokens	$15.00	$0.50
Context window	1,000,000	No data
Image input	Yes	No data
Developer	Moonshot AI	OpenAI

Charts

Partial coverage. Codex – GPT-6 Luna (high): 106 of 150 GeoBenchX tasks. Grey placeholders identify these entries in the index, cost and output token charts; their height does not represent a value. Comparable bars appear when first-attempt coverage is complete. Their available results remain in the side-by-side table.

Geospatial Agent Index

Equal-weight average of the evaluation scores, 0 to 100 · Higher is better

Attempts still to run

Data table: Geospatial Agent Index

Geospatial Agent Index
Model	Creator	Geospatial Agent Index	Range while attempts are pending	Benchmarks covered	Coverage (attempts)
Terminus-2 – Kimi K3 (always on)	Moonshot AI	72*	No pending judgements; coverage incomplete; 180 results awaiting review; 1 trace review declined by the judge	3 of 3	488 of 722 scheduled attempts

Cost per task

Average cost per task (USD) · Lower is better

Attempts still to run

How to read this chart

Not shown: partial coverage. 1 entries without a first attempt on every index task. What is still running is on the models page.

Data table: Cost per task

Cost per task
Model	Creator	Cost per task	Coverage (attempts)
Terminus-2 – Kimi K3 (always on)	Moonshot AI	$0.23	488 of 722 scheduled attempts

Output tokens per task

Average output tokens per attempt, reasoning included · Lower is better

Attempts still to run

How to read this chart

Not shown: partial coverage. 1 entries without a first attempt on every index task. What is still running is on the models page.

Data table: Output tokens per task

Output tokens per task
Model	Creator	Output tokens per task	Coverage (attempts)
Terminus-2 – Kimi K3 (always on)	Moonshot AI	6k	488 of 722 scheduled attempts

Frequently asked questions

Which scores higher on the Geospatial Agent Index, Terminus-2 – Kimi K3 (always on) or Codex – GPT-6 Luna (high)?

Not comparable yet: Codex – GPT-6 Luna (high) has partial coverage (106 of 150 GeoBenchX tasks), so it is not ranked. Terminus-2 – Kimi K3 (always on): 488 of 722 scheduled attempts. Codex – GPT-6 Luna (high): 430 of 722 scheduled attempts. Coverage is partial; scores can change as more attempts are decided.

Which is cheaper per task, Terminus-2 – Kimi K3 (always on) or Codex – GPT-6 Luna (high)?

Codex – GPT-6 Luna (high), at $0.0039 against $0.23 for Terminus-2 – Kimi K3 (always on).

Which writes fewer output tokens per task, Terminus-2 – Kimi K3 (always on) or Codex – GPT-6 Luna (high)?

Codex – GPT-6 Luna (high), at 3k against 6k for Terminus-2 – Kimi K3 (always on).

Which has the larger context window, Terminus-2 – Kimi K3 (always on) or Codex – GPT-6 Luna (high)?

Not known yet: Codex – GPT-6 Luna (high) has no context window data.

More: Terminus-2 – Kimi K3 (always on) · Codex – GPT-6 Luna (high) · all comparisons