STEMBench120 - Earth science: leaderboard
STEMBench120 - Earth science. v1.0.0; 120 tasks across six STEM domains; Codex tool runs distinct from closed-book rows; correctness-adjusted research progress, not answer accuracy. Reported metric: Research progress (0–4). Primary published table, with source configuration boundaries preserved.
Metric: Research progress (0–4). Source: www.ulam.ai. Saturation forecast: Around 2029. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 2.76 | #3 |
| 2 | GPT-6 Astra (xHigh) | 2.73 | #5 (GPT-6 Astra) |
| 3 | GPT-6.1 Sol | 2.68 | #8 |
| 4 | Muse Wonder | 2.66 | #16 |
| 5 | Gemini 4 Argon | 2.65 | #7 |
| 6 | MiMo-V2.6-Pro | 2.62 | #37 |
| 7 | Hy4 preview | 2.6 | #39 |
| 8 | DeepSeek V4.1 Flash | 2.59 | #36 |
| 9 | Mistral Large 4 | 2.49 | #98 |
No result here: #9 Claude Fable 5.1, #10 Claude Sonnet 5.5.
Interactive version: theaggregate.ai/benchmark?slug=stembench120-earth-science · How It Works · Data refreshed daily, snapshot 2026-10-16.