BenchBench-Protocol - Molecular Biology and Cloning: leaderboard
BenchBench-Protocol (Benchling, 2026) derives 149 wet-lab protocol-modification tasks from changes scientists made to published protocols, each graded against weighted rubric elements by a GPT-5.6 Terra judge. This board covers the 26 tasks from the molecular biology and cloning domain. Normalized rubric score in percent; higher is better.
Metric: Normalized rubric score (%; weighted rubric elements met by the response to real protocol-modification tasks in the molecular biology and cloning domain (26 tasks), graded by GPT-5.6 Terra; mean of repeated attempts per task). Source: arxiv.org. Saturation forecast: Around September 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (Max) | 64 | #20 (Claude Opus 5) |
| 2 | GPT-5.6 Sol (Max) | 51 | #29 (GPT-5.6 Sol) |
| 3 | GPT-5.6 Terra (Max) | 50.1 | #60 (GPT-5.6 Terra) |
| 4 | Kimi K3 (Max) | 49.6 | #53 (Kimi K3) |
| 5 | GPT-5.6 Luna (Max) | 46.1 | #117 (GPT-5.6 Luna) |
| 6 | Inkling (Max) | 46.1 | #270 (Inkling) |
| 7 | GLM-5.2 (Max) | 44.1 | #137 (GLM-5.2) |
| 8 | Grok 4.5 (High) | 42.2 | #78 (Grok 4.5) |
| 9 | Gemini 3.6 Flash (High) | 39.9 | #72 (Gemini 3.6 Flash) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=benchbench-protocol-molecular-biology-and-cloning · How It Works · Data refreshed daily, snapshot 2026-11-12.