Atlas

Benchmarks

← All benchmarks

MMLU-Bio (SecureBio)

Science

SecureBio's own runs of a 1,273-question biology subset of MMLU drawn from seven of its subject tasks. Evaluated in Inspect with zero-shot prompting, usually over 10 runs. SecureBio doesn't state how refusals enter accuracy. Claude Opus 4.8 prints 90.9% with 69.5% refused, more than it could score if refusals counted as wrong, so accuracy appears to cover answered questions only and each model's denominator differs. These scores aren't comparable with other runs of the test, which is why this is a diagnostic.

Benchmark source: Securebio

Top models (higher is better)

ModelScoreSource
Opus 4.793.8SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 4 Argon92.1SecureBio Biology Benchmark Dashboard · SecureBio
Opus 4.591.8SecureBio Biology Benchmark Dashboard · SecureBio
Grok 4.691.8SecureBio Biology Benchmark Dashboard · SecureBio
Sonnet 4.591.7SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 3.7 Flash91.4SecureBio Biology Benchmark Dashboard · SecureBio
Opus 4.191.3SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 3 Pro Preview91.2SecureBio Biology Benchmark Dashboard · SecureBio
Kimi K3.1 Pro91.2SecureBio Biology Benchmark Dashboard · SecureBio
Opus 491.1SecureBio Biology Benchmark Dashboard · SecureBio
Claude Opus 591.1SecureBio Biology Benchmark Dashboard · SecureBio
GPT-5.191.1SecureBio Biology Benchmark Dashboard · SecureBio
GPT-5.291.1SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 2.5 Pro91.0SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 3.6 Flash91.0SecureBio Biology Benchmark Dashboard · SecureBio
GPT-591.0SecureBio Biology Benchmark Dashboard · SecureBio
GPT-5 Pro91.0SecureBio Biology Benchmark Dashboard · SecureBio
Grok 4.1 Fast91.0SecureBio Biology Benchmark Dashboard · SecureBio
Opus 4.890.9SecureBio Biology Benchmark Dashboard · SecureBio
Sonnet 590.9SecureBio Biology Benchmark Dashboard · SecureBio
o1 (2024-12-17)90.9SecureBio Biology Benchmark Dashboard · SecureBio
o390.9SecureBio Biology Benchmark Dashboard · SecureBio
Qwen3.8 2.4T A95B90.9SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 3 Flash Preview90.8SecureBio Biology Benchmark Dashboard · SecureBio
GPT-5.6 Terra90.7SecureBio Biology Benchmark Dashboard · SecureBio
Grok 4 Fast90.7SecureBio Biology Benchmark Dashboard · SecureBio
o4-mini90.7SecureBio Biology Benchmark Dashboard · SecureBio
Qwen3 235B A22B Thinking 250790.7SecureBio Biology Benchmark Dashboard · SecureBio
Sonnet 490.6SecureBio Biology Benchmark Dashboard · SecureBio
GPT-4.590.5SecureBio Biology Benchmark Dashboard · SecureBio
Grok 490.5SecureBio Biology Benchmark Dashboard · SecureBio
Gemini 3.1 Pro Preview90.2SecureBio Biology Benchmark Dashboard · SecureBio
Qwen3.8 27B90.1SecureBio Biology Benchmark Dashboard · SecureBio
Haiku 4.589.9SecureBio Biology Benchmark Dashboard · SecureBio
Qwen3-Next-80B-A3B-Thinking89.9SecureBio Biology Benchmark Dashboard · SecureBio
DeepSeek-R1-052889.8SecureBio Biology Benchmark Dashboard · SecureBio
GPT-4.189.7SecureBio Biology Benchmark Dashboard · SecureBio
Llama 4 Maverick Instruct FP889.6SecureBio Biology Benchmark Dashboard · SecureBio
Mistral Medium 3.589.4SecureBio Biology Benchmark Dashboard · SecureBio
Kimi K2 Instruct89.3SecureBio Biology Benchmark Dashboard · SecureBio
Loading Atlas data…