Axis Spatial

Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) vs. Codex – GPT-6 Luna (xhigh): geospatial agent comparison

Comparison between Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) and Codex – GPT-6 Luna (xhigh) 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 – Mistral Small 3.1 24B (Non-reasoning) vs. Codex – GPT-6 Luna (xhigh)
MetricTerminus-2 – Mistral Small 3.1 24B (Non-reasoning)Codex – GPT-6 Luna (xhigh)
Geospatial Agent Index0* (attempts per task: 1 on 9 of 290 tasks so far)77* (attempts per task: 2.4)
GeoAgentBench scoreNo data99
GeoBenchX scoreNo data61
Earth-Bench score054
GeoAnalystBench scoreNo data95
Raster and remote sensing score075
Vector and overlay scoreNo data78
Networks and routing scoreNo data100
Climate and time series score074
Spatial statistics and interpolation scoreNo data65
Mapping and cartography scoreNo data64
Recognising infeasible requests scoreNo data61
Cost per task$0.247$0.0092
Time per task15.6 min3.4 min
Tokens per attempt693k233k
Turns per attempt25.79.7
Attempts decided9627
Input price per 1M tokens$0.35$0.10
Output price per 1M tokens$0.56$0.50
Context window128,000No data
Image inputYesNo data
DeveloperMistral AIOpenAI

Charts

Geospatial Agent Index

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

  • Partial coverage
Geospatial Agent IndexCodex – GPT-6 Luna (xhigh): 77* (partial coverage); Terminus-2 – Mistral Small 3.1 24B (Non-reasoning): 0* (partial coverage)025507510077*GPT-6 Luna (xhigh)0*Mistral Small 3.1 24B (Non-reasoning)
Data table: Geospatial Agent Index
Geospatial Agent Index
ModelCreatorGeospatial Agent IndexRange while attempts are pendingBenchmarks coveredCoverage (attempts)
Codex – GPT-6 Luna (xhigh)OpenAI77* (attempts per task: 2.4)69 to 784 of 4627 of 870 planned attempts
Terminus-2 – Mistral Small 3.1 24B (Non-reasoning)Mistral AI0* (attempts per task: 1 on 9 of 290 tasks so far)No pending judgements; coverage incomplete1 of 49 of 351 planned attempts

Cost per task

Average cost per task (USD) · Lower is better

  • Partial coverage
Cost per taskCodex – GPT-6 Luna (xhigh): $0.0092 (partial coverage); Terminus-2 – Mistral Small 3.1 24B (Non-reasoning): $0.247 (partial coverage)$0$0.062$0.125$0.188$0.250$0.0092GPT-6 Luna (xhigh)$0.247Mistral Small 3.1 24B (Non-reasoning)
Data table: Cost per task
Cost per task
ModelCreatorCost per taskCoverage (attempts)
Terminus-2 – Mistral Small 3.1 24B (Non-reasoning)Mistral AI$0.2479 of 351 planned attempts
Codex – GPT-6 Luna (xhigh)OpenAI$0.0092706 of 870 planned attempts

Time per task

Average agent wall time per task · Lower is better

  • Partial coverage
Time per taskCodex – GPT-6 Luna (xhigh): 3.4 min (partial coverage); Terminus-2 – Mistral Small 3.1 24B (Non-reasoning): 15.6 min (partial coverage)0m4m8m12m17m3.4 minGPT-6 Luna (xhigh)15.6 minMistral Small 3.1 24B (Non-reasoning)
Data table: Time per task
Time per task
ModelCreatorTime per taskCoverage (attempts)
Terminus-2 – Mistral Small 3.1 24B (Non-reasoning)Mistral AI15.6 min9 of 351 planned attempts
Codex – GPT-6 Luna (xhigh)OpenAI3.4 min706 of 870 planned attempts

Frequently asked questions

Which scores higher on the Geospatial Agent Index, Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) or Codex – GPT-6 Luna (xhigh)?
Not comparable yet: Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) has scores on 1 of 4 benchmarks, so it is not yet ranked. Terminus-2 – Mistral Small 3.1 24B (Non-reasoning): 9 of 351 planned attempts. Codex – GPT-6 Luna (xhigh): 627 of 870 planned attempts. Coverage is partial; scores can change as more attempts are decided.
Which is cheaper per task, Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) or Codex – GPT-6 Luna (xhigh)?
Codex – GPT-6 Luna (xhigh), at $0.0092 against $0.247 for Terminus-2 – Mistral Small 3.1 24B (Non-reasoning).
Which is faster per task, Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) or Codex – GPT-6 Luna (xhigh)?
Codex – GPT-6 Luna (xhigh), at 3.4 min against 15.6 min for Terminus-2 – Mistral Small 3.1 24B (Non-reasoning).
Which has the larger context window, Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) or Codex – GPT-6 Luna (xhigh)?
Not known yet: Codex – GPT-6 Luna (xhigh) has no context window data.

More: Terminus-2 – Mistral Small 3.1 24B (Non-reasoning) · Codex – GPT-6 Luna (xhigh) · all comparisons