SynBio-Reason (Zero-Shot): leaderboard
Genetic circuit design as code generation (SynBio-Reason, Flynn, 2026): the model writes pysbol3 code that builds the requested circuit in SBOL3, and a five-level verifier checks execution, validity, structure, semantics and function. Task success rate in percent, averaged over procedural, Cello and literature circuit splits, with a zero-shot prompt; higher is better.
Metric: Task success rate (%; mean over the procedural test, Cello in-distribution, Cello held-out-part and Literature-91 de novo design splits; generated pysbol3 code must pass all five verification levels; zero-shot prompt with the pysbol3 API reference, greedy decoding). Source: arxiv.org. Saturation forecast: Around March 2028. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 (Thinking) | 22.4 | #152 (Claude Opus 4.5) |
| 2 | Qwen 3 8B | 7 | #1838 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=synbio-reason-zero-shot · How It Works · Data refreshed daily, snapshot 2026-11-12.