Skip to content

Benchmark

Protein design · Library ranking

Science benchmarks measure how well models answer science questions. Scores are percentages. Higher is better.

Models measured
5
Scores recorded
8
Sources
2
IQ weight
0.63

Score distribution

49.7%
Mean
53.5%
Median
5.1%
Std dev
44%
10th percentile
54.8%
90th percentile
44%
Min
44%56%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.63 for chat models, measured over 5 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 5.5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56%
2Claude Sonnet 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim54.8%
3Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.54.8%
4Claude Opus 5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.5%
5Claude Haiku 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim45.3%
6Claude Haiku 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.45.3%
7Claude Sonnet 5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.44%
8Claude Sonnet 5Opus 5.5 System Card · adaptive thinking · max effort unless stated· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.44%

Data credits: 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. · Data by LLM Stats