ProteinGeoBench - Selected 20 Tasks: leaderboard
ProteinGeoBench (2026): 26 models on 20 discrimination-selected tasks, 2048 graded trials, up to four runs per task; temperature 1 where accepted and no explicit reasoning setting. Different tasks and sampling from the main four-model comparison; no cross-table score pooling. Mean Reward (0–1); higher is better.
Metric: Mean Reward (0–1). Source: arxiv.org. Saturation forecast: Around March 2028. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 0.6 | #12 |
| 2 | Gemini 3.5 Flash | 0.58 | #60 |
| 3 | Muse Spark 1.1 | 0.55 | #52 |
| 4 | Claude Sonnet 5 | 0.54 | #56 |
| 5 | Gemini 3.7 Flash | 0.51 | #21 |
| 6 | Grok 4.5 | 0.49 | #47 |
| 7 | Claude Opus 4.8 | 0.49 | #37 |
| 8 | GPT-5.6 Sol | 0.48 | #18 |
| 9 | Kimi K3 | 0.48 | #31 |
| 10 | GPT-5.6 Terra | 0.46 | #41 |
| 11 | Kimi K2.7 Code | 0.45 | #89 |
| 12 | GPT-5.5 Pro | 0.44 | #17 |
| 13 | GPT-5.5 | 0.44 | #30 |
| 14 | DeepSeek V4 Pro | 0.43 | #101 |
| 15 | Gemini 3.1 Pro (Preview) | 0.42 | #58 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=proteingeobench-selected-20-tasks · How It Works · Data refreshed daily, snapshot 2026-10-21.