MMLU

Massive Multitask Language Understanding — 57 subjects spanning STEM, humanities, and social sciences.

Reasoning Multilingual

Knowledge Q&A General assistants Exam-style reasoning

Higher is better · accuracy %

1 Claude Opus 5 89.2%

2 Gemini 2.5 Pro 88.6%

3 Claude Sonnet 5 86.1%

4 DeepSeek-V4 Flash 85.4%

5 GPT OSS 120B 83%

Bars scaled from 75 so close scores stay legible — the figure at right is the actual accuracy %.