SciUniverse: leaderboard
C5R's benchmark of hands-on scientific work: models direct chemistry, biology and materials experiments at the Facility-0 lab, controlling instruments and instructing human operators, or in its digital twin. Level 1 has 92 tasks in 17 families, from powder pressing, synthesis and liquid-handler control to NMR, XRD and LC-MRM interpretation, reaction optimization under a budget and two simulated weeks of lab management. Pass@1, with each family weighted equally.
Metric: Pass@1 (%). Source: c5r.net. Saturation forecast: Around March 2028. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (xHigh) | 45.3 | #14 (Claude Fable 5.1) |
| 2 | GPT-6 Astra (xHigh) | 32.5 | #6 (GPT-6 Astra) |
| 3 | Claude Opus 5 (xHigh) | 30.5 | #20 (Claude Opus 5) |
| 4 | Grok 4.6 (xHigh) | 26.2 | #44 (Grok 4.6) |
| 5 | Gemini 3.8 Flash (High) | 14.6 | #36 (Gemini 3.8 Flash) |
| 6 | GPT-5.6 Sol (xHigh) | 9.4 | #29 (GPT-5.6 Sol) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sciuniverse · How It Works · Data refreshed daily, snapshot 2026-11-12.