LabBench (Gamow Labs): leaderboard
LabBench (Gamow Labs, 2026) — 20 held-out wet-lab decision tasks built from real drug-discovery and genomics lab records. The agent receives the records that existed at a decision point, with the lab's interpretation withheld, and must commit to the next step; it is graded on 20-22 binary criteria per task tied to what the lab actually decided. Mean share of criteria passed. Unrelated to FutureHouse's LAB-Bench.
Metric: Mean share of binary rubric criteria passed (%) over 20 held-out wet-lab decision tasks (406 criteria), one run per task in each vendor's harness, judged by Codex with GPT-5.6 Sol. Source: gamowlabs.com. Saturation forecast: Around 2028. 5 models tracked. Updated 1 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6 Astra | 40.8 | #6 | |
| 2 | Claude Opus 5.5 | Max | 40.5 | #5 |
| 3 | Grok 4.7 | 29.6 | #47 | |
| 4 | Muse Spark 1.3 | High | 28.2 | #43 |
| 5 | Gemini 3.8 Flash | 18.4 | #36 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #10 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=labbench-gamow-labs · How It Works · Data refreshed daily, snapshot 2026-11-16.