BioMysteryBench

Solving real bioinformatics mysteries from raw data as an agent. Run by Vals AI.

CategoryReasoning
MetricAccuracy (%)
SourceVals AI (evaluator)
LicenseNone stated
Benchmark pagewww.vals.ai
StatusScored
Weight0.9 within Reasoning
FitDiscrimination 0.67, noise 0.33

Results

#ModelProviderResultDate
1 Claude Fable 5.2 High Anthropic 83.7% 2026-11-09
2 Claude Sonnet 5.5 Max Anthropic 81.1% 2026-11-09
3 GPT-6.1 Sol Max OpenAI 79.6% 2026-11-09
4 Claude Opus 5.5 Max Anthropic 79.3% 2026-11-09
5 GPT-6 Astra Max OpenAI 79.3% 2026-11-09
6 Claude Opus 5 Max Anthropic 79.3% 2026-11-09
7 Gemini 4 Argon High Google 76.3% 2026-11-09
8 Kimi K3.1 Pro Max Moonshot AI 75.6% 2026-11-09
9 GPT-6 Sol Max OpenAI 74.8% 2026-11-09
10 Muse Wonder Full Max Meta 73.3% 2026-11-09
11 Kimi K3 Max Moonshot AI 72.6% 2026-11-09
12 Grok 4.6 High xAI 72.2% 2026-11-09
13 GPT-5.6 Sol Max OpenAI 71.1% 2026-11-09
14 Claude Haiku 5.5 Max Anthropic 70.4% 2026-11-09
15 Inkling 2 Small Thinking Machines 70.0% 2026-11-09
16 Hy4 preview Tencent 69.3% 2026-11-09
17 MiMo-V2.6-Flash Xiaomi 69.3% 2026-11-09
18 Grok 4.7 Extra High xAI 69.3% 2026-11-09
19 DeepSeek V4.1 Flash High DeepSeek 67.8% 2026-11-09
20 Step 5 Preview StepFun 67.4% 2026-11-09
21 Ember-1 Fireworks 67.0% 2026-11-09
22 Mistral Large 4 High Mistral 67.0% 2026-11-09
23 Muse Wonder Max Meta 67.0% 2026-11-09
24 Muse Spark 1.2 Extra High Meta 64.8% 2026-11-09
25 DeepSeek V4 Flash 0731 High DeepSeek 64.4% 2026-11-09
26 Gemini 3.8 Flash High Google 62.2% 2026-11-09
27 GPT-6 Luna Max OpenAI 61.5% 2026-11-09
28 GPT-5.6 Luna Max OpenAI 61.5% 2026-11-09
29 Gemini 3.1 Pro Preview High Google 60.0% 2026-11-09
30 Gemini 3.6 Flash High Google 58.5% 2026-11-09
31 Beam 1.0 Reflection 42.6% 2026-11-09