Skip to content

Benchmark

BioMysteryBench

Scores are percentages. Higher is better.

Models measured
12
Scores recorded
30
Sources
2
IQ weight
0.00

Score distribution

66.1%
Mean
71.3%
Median
21.3%
Std dev
40%
10th percentile
89.3%
90th percentile
20.9%
Min
20.9%91.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.00 for chat models, measured over 12 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Solvable · 73 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.91.4%
2Claude Opus 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim90.1%
3Claude Opus 5Anthropic Opus 5 launch table · adaptive thinking · max effort unless noted · human-solvable subset· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.1%
4Claude Opus 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.3%
5Claude Opus 5.5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Solvable · 73 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.3%
6Claude Sonnet 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.2%
7Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · Human Solvable subset· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.89.2%
8Gemini 3.8 FlashHuman Solvable subset· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.8%
9Gemini 3.7 FlashHuman Solvable subset· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87.1%
10Claude Sonnet 5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Solvable · 73 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.9%
11Claude Sonnet 5human-solvable subset · source plots 0.82· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82%
12Kimi K3.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.6%
13Claude Opus 4.8Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · Human· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.4%
14Claude Opus 4.7Verified subset · bash/package tools · thinking disabled· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.9%
15Hy4 preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim71.3%
16Hy4 previewofficial model-card appendix· Reported by Tencent Hy· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.3%
17Gemini 3.8 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim56.5%
18Gemini 3.8 FlashHuman Difficult subset· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.5%
19Claude Opus 5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Difficult · 17 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.8%
20Claude Sonnet 4.6bash/code/package tools · no extended thinking· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.50.4%
21Claude Opus 5.5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Difficult · 17 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.50%
22Claude Opus 5Anthropic Opus 5 launch table · adaptive thinking · max effort unless noted · hard subset· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.49.4%
23Claude Fable 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim46.1%
24Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · Human Difficult subset· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.44.7%
25Gemini 3.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim43.5%
26Gemini 3.7 FlashHuman Difficult subset· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43.5%
27Claude Opus 4.8Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · Hard· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.40%
28Claude Sonnet 5Opus 5.5 System Card · adaptive thinking · max effort unless stated · Human Difficult · 17 tasks· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.39.4%
29Claude Sonnet 5human-difficult subset · source plots 0.35· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.35%
30Claude Opus 4.7Hard subset · objective ground truth · unsolved by human experts · bash/package tools · thinking disabled· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.20.9%

Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.