BioPhys-Bridge: leaderboard
Interdisciplinary reasoning over physics-grounded biological research, scored on whether the answer identifies the evidence it rests on rather than only on the answer text. Evidence-ID F1, 0-1, higher is better. Source: arxiv.org/abs/2609.19180.
Metric: Evidence-ID F1 (0-1). Source: arxiv.org. Saturation forecast: Not forecast. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Flash | 0.36 | #155 |
| 2 | Qwen 3.7 Max | 0.32 | #85 |
| 3 | GPT-4o Mini | 0.29 | #598 |
| 4 | Claude Opus 4.8 | 0.24 | #39 |
| 5 | Claude Sonnet 5 | 0.22 | #65 |
| 6 | DeepSeek V4 Pro | 0.11 | #112 |
| 7 | Gemini 2.5 Pro | 0.01 | #159 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=biophys-bridge · How It Works · Data refreshed daily, snapshot 2026-11-05.