Gemma 4 31B (Reasoning) vs gpt-oss-120b (high)
Results updated 11 November 2026 · In Terminus-2 and Codex, so part of any gap can come from the harness · How we benchmark
Verdict
Gemma 4 31B (Reasoning) clearly beats gpt-oss-120b (high) by 12 points, and is cheaper and faster.
- Cost per task
- $0.015 vs $0.26
- Time per task
- 2.1 min vs 6.0 min
- Tokens per attempt
- 93k vs 700k
- Image input
- Yes vs No
Gemma 4 31B (Reasoning), Terminus-2 scores higher on GeoBenchX and GS-QA2; gpt-oss-120b (high), Codex scores higher on Earth-Bench V2. Gemma 4 31B (Reasoning), Terminus-2 is cheaper and faster: $0.015 against $0.26 per task, and 2.1 min against 6.0 min.
Where they differ most
By kind of work, share of tasks passed, 0 to 100 · Higher is better
| Kind of work | Gemma 4 31B (Reasoning) | gpt-oss-120b (high) |
|---|---|---|
| GIS analysis | 56 | 23 |
| Remote sensing and raster analysis | 59 | 38 |
| Making maps | 70 | 61 |
| Change detection | 52 | 60 |
Where the tasks are set: each kind of work links to its tasks and scoresCharts for each: Gemma 4 31B (Reasoning), Terminus-2 · gpt-oss-120b (high), Codex.
Side by side
| Metric | Gemma 4 31B (Reasoning), Terminus-2 | gpt-oss-120b (high), Codex |
|---|---|---|
| Geospatial Agent Index: Text only | 55 | 43 |
| Geospatial Agent Index: Text + vision | No data | No image input |
| Earth-Bench V2 score | 44 | 49 |
| GeoBenchX score | 61 | 55 |
| GS-QA2 score | 61 | 25 |
| Vision agents score | No data | No image input |
| GeoAgentBench score (Guided workflows; not in the Agent Index) | 83 | 58 |
| GeoAnalystBench score (Guided workflows; not in the Agent Index) | 63 | 42 |
| OGD4All score (standalone; in no index) | 85 | No data |
| GeoNatureAgent score (standalone; in no index) | 100 | No data |
| MultiGlobeQA score (standalone; in no index) | 30 | No data |
| GeoAgentEval score (standalone; in no index) | 88 | No data |
| ImpactMesh score (standalone; in no index) | No data | No data |
| Remote sensing and raster analysis score | 59 | 38 |
| Change detection score | 52 | 60 |
| GIS analysis score | 56 | 23 |
| Climate and time series score | 47 | 40 |
| Making maps score | 70 | 61 |
| Cost per task | $0.015 | $0.26 |
| Time per task (at 100 output tokens/s) | 2.1 min | 6.0 min |
| Tokens per attempt | 93k | 700k |
| Turns per attempt | 9.0 | 29.8 |
| Input price per 1M tokens | $0.17 | $0.15 |
| Output price per 1M tokens | $0.40 | $0.59 |
| Context window | 262,144 | 128,000 |
| Image input | Yes | No |
| Developer | OpenAI |
Frequently asked questions
Which scores higher on the Geospatial Agent Index, Gemma 4 31B (Reasoning), Terminus-2 or gpt-oss-120b (high), Codex?
Gemma 4 31B (Reasoning), Terminus-2, 55 to 43.
Which is cheaper per task, Gemma 4 31B (Reasoning), Terminus-2 or gpt-oss-120b (high), Codex?
Gemma 4 31B (Reasoning), Terminus-2, $0.015 to $0.26.
Which writes fewer output tokens per task, Gemma 4 31B (Reasoning), Terminus-2 or gpt-oss-120b (high), Codex?
Gemma 4 31B (Reasoning), Terminus-2, 8k to 27k.
Which has the larger context window, Gemma 4 31B (Reasoning), Terminus-2 or gpt-oss-120b (high), Codex?
Gemma 4 31B (Reasoning), Terminus-2, 262,144 tokens to 128,000 tokens.