BioMysteryBench Human Difficult (BioMysteryBench (human-difficult))
We show this table for reference; we do not rank on it.
Computational biology challenges with objective answers that remained unsolved by independent human experts.
Benchmark score on BioMysteryBench (human-difficult) — November 13, 2026
We compile the BioMysteryBench (human-difficult) rows from provider self-reports. Gemini 3.8 Flash leads the table at 56.5%, followed by Claude Opus 5.5 (50.0%) and Claude Opus 5 (49.4%). We do not use these results to rank models overall.
Gemini 3.8 Flash
Claude Opus 5.5
Anthropic
Reported by AnthropicClaude Opus 5
Anthropic
Reported by Anthropic6 modelsKnowledgeCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (6 models)
All 6 results on this page were reported by the labs that built the models.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Gemini 3.8 FlashGoogleReported by Google | 56.5% | Reported by Google | Not reported | Closed |
| 2 | Claude Opus 5.5AnthropicReported by Anthropic | 50.0% | Reported by Anthropic | Not reported | Closed |
| 3 | Claude Opus 5AnthropicReported by Anthropic | 49.4% | Reported by Anthropic | Not reported | Closed |
| 4 | Claude Sonnet 5.5AnthropicReported by Anthropic | 44.7% | Reported by Anthropic | Not reported | Closed |
| 5 | Gemini 3.7 FlashGoogleReported by Google | 43.5% | Reported by Google | Not reported | Closed |
| 6 | Apodex 1.1ApodexReported by Apodex | 35.3% | Reported by Apodex | Not reported | Closed |
- Lab-reported
- Benchmark-run
- Independent
Results reported by model developersUnranked
Labs ran these themselves, with a metric, split or setting the table above does not use. They take no rank here.
| Model | Reported value | What was measured | Against the table above | Who ran it |
|---|---|---|---|---|
| Apodex 1.1Apodex | 23.5%Tasks solved (4 of 17) | ReAct single agentThe stored row uses Agent Team; revised 17-task Human-difficult Set | Not comparable to the scores above | Reported by Apodex |
A lab’s own result is shown for reference. It never takes a rank on this board.
How we label sourcesAmong the reported BioMysteryBench (human-difficult) rows, Gemini 3.8 Flash is first at 56.5%. The third row is 7.1 points behind. The broader top-10 range is 21.2 points, so the table still separates the published systems.
6 models have been evaluated on BioMysteryBench (human-difficult). The benchmark falls in the Knowledge category. BioMysteryBench (human-difficult) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About BioMysteryBench (human-difficult)
Year
2026
Tasks
Human-difficult computational biology investigations
Format
Task score
Difficulty
Frontier computational biology
Section 8.17.1 reports the provider's revised human-difficult subset after removing six problems based on external feedback.
Freshness and provenance
Version
BioMysteryBench (human-difficult) 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does BioMysteryBench (human-difficult) measure?
Computational biology challenges with objective answers that remained unsolved by independent human experts.
Which model scores highest on BioMysteryBench (human-difficult)?
Gemini 3.8 Flash by Google currently leads with a score of 56.5% on BioMysteryBench (human-difficult).
How many models are evaluated on BioMysteryBench (human-difficult)?
6 AI models have published results on BioMysteryBench (human-difficult) in the BenchLM catalog.
Compare top models on BioMysteryBench (human-difficult)
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.