BioMysteryBench
Claude Opus 5 90.1% leads since Jul 2026 In use
Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits.
Version v11 (2026-07-06) removed 9 problems with wrong or underivable keys and edited 24, leaving 90 (73 human-solvable, 17 human-hard). Cards report the two splits apart. A 5-problem preview is open and the full set is gated.
Made by Anthropic Released Apr 2026 Lab and data work All tasks public Last in a launch post Oct 2026
Last model added Oct 2026
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
8
Models scored in lab cards
6
Launch posts list it, from 4 labs
10
System cards report it, from 2 labs
Claude Opus 5 leads with 90.1%
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
- 1 Claude Opus 5 90.1%
- 2
Gemini 3.8 Flash 88.8% - 3 Gemini 3.7 Flash 87.1%
- 4 Claude Fable 5 83.9%
- 5 Kimi K3.1 Pro 80.6%
- 6
Tencent Hy4 preview 71.3% - 7 Claude Opus 4.6 61.5%
- 8 Claude Sonnet 4.6 50.4%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 90.1% | Jul 2026 | Model card | |
| 2 | Gemini 3.8 Flash | 88.8% | Sep 2026 | Model card | |
| 3 | Gemini 3.7 Flash | 87.1% | Aug 2026 | Model card | |
| 4 | Claude Fable 5 | 83.9% | Jun 2026 | Model card | |
| 5 | Kimi K3.1 Pro | 80.6% | Oct 2026 | Model card | |
| 6 | Tencent Hy4 preview | 71.3% | Aug 2026 | Model card | |
| 7 | Claude Opus 4.6 | 61.5% | Feb 2026 | Model card | |
| 8 | Claude Sonnet 4.6 | 50.4% | Feb 2026 | Model card |
3 models have held number one since its release in Apr 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Anthropic
The best score passed 85% in Jul 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Reported by the lab in its model card
Best so far
4 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 24 Oct 2026 | Kimi K3.1 Pro | 80.6 | BioMysteryBench (v11) |
| 2 Sep 2026 | Gemini 3.8 Flash | 88.8 | Best of BioMysteryBench, Human Solvable / BioMysteryBench, Human Difficult |
| 28 Aug 2026 | Tencent Hy4 preview | 71.3 | BioMysteryBench |
| 13 Aug 2026 | Gemini 3.7 Flash | 87.1 | Best of BioMysteryBench, Human solvable / BioMysteryBench, Human difficult |
| 24 Jul 2026 | Claude Opus 5 | 90.1 | Best of Biology BioMysteryBench (hard) / Biology BioMysteryBench (human solved) |
| 9 Jun 2026 | Claude Fable 5 | 83.9 | Best of Biology BioMysteryBench (hard) / Biology BioMysteryBench (human solved) |
10 system cards report it, from 2 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 28 Sep 2026 | Claude Sonnet 5.5 system card | BioMysteryBench, Human Solvable 89.2% and Human Difficult 44.7%, no overall (2.2.2 and 8.17.1, Figure 8.17.A) | Bio and chem risk | |
| 22 Sep 2026 | Claude Opus 5.5 system card | BioMysteryBench, Human Solvable 89.3% and Human Difficult 50.0%, no overall, final model (2.2.3.1 and 8.17.1) | Bio and chem risk | |
| 2 Sep 2026 | Gemini 3.8 Flash model card | 88.8 | BioMysteryBench, Human Solvable (results table, p.5). Human Difficult = 56.5% | Capabilities |
| 13 Aug 2026 | Gemini 3.7 Flash model card | 87.1 | BioMysteryBench, Human solvable (results table, p.5). Human difficult = 43.5% | Capabilities |
| 24 Jul 2026 | Claude Opus 5 system card | BioMysteryBench, Human Solvable 90.1% and Human Difficult 49.4%, no overall (8.17.1) | Capabilities | |
| 30 Jun 2026 | Claude Sonnet 5 system card | BioMysteryBench, Human Solvable 0.82 and Human Difficult 0.35, no overall (8.14.1, Figure 8.14.6.A) | Capabilities | |
| 28 May 2026 | Claude Opus 4.8 system card | BioMysteryBench Verified, Human Solvable 80.4% and Human Difficult 40.0%, no overall (8.16.2) | Capabilities | |
| 16 Apr 2026 | Claude Opus 4.7 system card | BioMysteryBench, Verified (human-solvable) subset 78.9 and Hard subset 20.9, no overall, no extended thinking (8.13.1.2) | Capabilities | |
| 17 Feb 2026 | Claude Sonnet 4.6 system card | 50.4 | Computational Biology, BioMysteryBench, no extended thinking (2.21.1) | Capabilities |
| 5 Feb 2026 | Claude Opus 4.6 system card | 61.5 | Computational Biology, BioMysteryBench, bash tool, no extended thinking, above the human expert baseline (2.22, 8.2.4.10) | Capabilities |
More in Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Argon 96.4%
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87% - HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
- SGI-Bench In use Models work through more than 1,000 expert samples in four scientist tasks: deep research, idea generation, dry and wet experiments, and experimental reasoning. Intern-S2-397B 51.9%