Terminus-2 – Qwen 3.8 27B (xhigh) vs. Codex – GPT-5.6 Luna (low): geospatial agent comparison

Comparison between Terminus-2 – Qwen 3.8 27B (xhigh) and Codex – GPT-5.6 Luna (low) 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 – Qwen 3.8 27B (xhigh) vs. Codex – GPT-5.6 Luna (low)
Metric	Terminus-2 – Qwen 3.8 27B (xhigh)	Codex – GPT-5.6 Luna (low)
Geospatial Agent Index	37* (partial: attempt 1 on 27 of 488 tasks)	60* (partial: attempt 1 on 397 of 488 tasks)
Earth-Bench V2 score	37	44
GeoBenchX score	No data	61
GS-QA2 score	No data	76
GeoAgentBench score (guided GIS workflow; outside the index)	60	86
GeoAnalystBench score (guided GIS workflow; outside the index)	100	95
Raster and remote sensing score	38	70
Vector and overlay score	No data	72
Climate and time series score	36	42
Spatial statistics and interpolation score	No data	0
Mapping and cartography score	No data	83
Recognising infeasible requests score	No data	50
Cost per task	$0.055	$0.0057
Time per task	19.1 min	1.6 min
Tokens per attempt	66k	84k
Turns per attempt	7.0	7.1
Attempts decided	27	397
Input price per 1M tokens	$0.45	$0.20
Output price per 1M tokens	$3.20	$1.20
Context window	262,144	No data
Image input	Yes	No data
Developer	Alibaba (Qwen)	OpenAI

Charts

Partial coverage. Terminus-2 – Qwen 3.8 27B (xhigh): 27 of 117 Earth-Bench V2 tasks; GeoBenchX incomplete; GS-QA2 incomplete; Codex – GPT-5.6 Luna (low): 59 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

Data table: Geospatial Agent Index

Geospatial Agent Index
Model	Creator	Geospatial Agent Index	Range while attempts are pending	Benchmarks covered	Coverage (attempts)

Cost per task

Average cost per task (USD) · Lower is better

How to read this chart

Not shown: partial coverage. 2 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)

Output tokens per task

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

How to read this chart

Not shown: partial coverage. 2 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)

Frequently asked questions

Which scores higher on the Geospatial Agent Index, Terminus-2 – Qwen 3.8 27B (xhigh) or Codex – GPT-5.6 Luna (low)?

Not comparable yet: Terminus-2 – Qwen 3.8 27B (xhigh) has partial coverage (27 of 117 Earth-Bench V2 tasks; GeoBenchX incomplete; GS-QA2 incomplete) and Codex – GPT-5.6 Luna (low) has partial coverage (59 of 150 GeoBenchX tasks), so it is not ranked. Terminus-2 – Qwen 3.8 27B (xhigh): 27 of 501 scheduled attempts. Codex – GPT-5.6 Luna (low): 397 of 722 scheduled attempts. Coverage is partial; scores can change as more attempts are decided.

Which is cheaper per task, Terminus-2 – Qwen 3.8 27B (xhigh) or Codex – GPT-5.6 Luna (low)?

Codex – GPT-5.6 Luna (low), at $0.0057 against $0.055 for Terminus-2 – Qwen 3.8 27B (xhigh).

Which writes fewer output tokens per task, Terminus-2 – Qwen 3.8 27B (xhigh) or Codex – GPT-5.6 Luna (low)?

Codex – GPT-5.6 Luna (low), at 1k against 9k for Terminus-2 – Qwen 3.8 27B (xhigh).

Which has the larger context window, Terminus-2 – Qwen 3.8 27B (xhigh) or Codex – GPT-5.6 Luna (low)?

Not known yet: Codex – GPT-5.6 Luna (low) has no context window data.

More: Terminus-2 – Qwen 3.8 27B (xhigh) · Codex – GPT-5.6 Luna (low) · all comparisons