Benchmark
Vals BioMysteryBench
Science benchmarks measure how well models answer science questions. Scores are percentages.
- Models measured
- 33
- Scores recorded
- 33
- Sources
- 1
- IQ weight
- 2.27
Score distribution
- 70.2%
- Mean
- 70%
- Median
- 8.2%
- Std dev
- 61.5%
- 10th percentile
- 79.6%
- 90th percentile
- 42.6%
- Min
This benchmark has a weight of 2.27 for chat models, measured over 33 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.2· Vals AI | 83.7% |
| 2 | Claude Sonnet 5.5· Vals AI | 81.1% |
| 3 | GPT-6.2 Astra· Vals AI | 80.7% |
| 4 | GPT-6.1 Sol· Vals AI | 79.6% |
| 5 | Claude Opus 5· Vals AI | 79.3% |
| 6 | Claude Opus 5.5· Vals AI | 79.3% |
| 7 | GPT-6 Astra· Vals AI | 79.3% |
| 8 | GPT-6.2 Sol· Vals AI | 77.8% |
| 9 | Gemini 4 Argon· Vals AI | 76.3% |
| 10 | Kimi K3.1 Pro· Vals AI | 75.6% |
| 11 | GPT-6 Sol· Vals AI | 74.8% |
| 12 | Muse Wonder Full· Vals AI | 73.3% |
| 13 | Kimi K3· Vals AI | 72.6% |
| 14 | Grok 4.6· Vals AI | 72.2% |
| 15 | GPT-5.6 Sol· Vals AI | 71.1% |
| 16 | Claude Haiku 5.5· Vals AI | 70.4% |
| 17 | Inkling 2 Small· Vals AI | 70% |
| 18 | Grok 4.7· Vals AI | 69.3% |
| 19 | Hy4 preview· Vals AI | 69.3% |
| 20 | MiMo V2.6 Flash· Vals AI | 69.3% |
| 21 | DeepSeek V4.1 Flash· Vals AI | 67.8% |
| 22 | Step 5 Preview· Vals AI | 67.4% |
| 23 | Ember-1· Vals AI | 67% |
| 24 | Mistral Large 4· Vals AI | 67% |
| 25 | Muse Wonder· Vals AI | 67% |
| 26 | Muse Spark 1.2· Vals AI | 64.8% |
| 27 | DeepSeek V4 Flash 0731· Vals AI | 64.4% |
| 28 | Gemini 3.8 Flash· Vals AI | 62.2% |
| 29 | GPT-5.6 Luna· Vals AI | 61.5% |
| 30 | GPT-6 Luna· Vals AI | 61.5% |
| 31 | Gemini 3.1 Pro Preview· Vals AI | 60% |
| 32 | Gemini 3.6 Flash· Vals AI | 58.5% |
| 33 | Beam 1.0 501B A23B· Vals AI | 42.6% |