List your company

Language Understanding on MMLU (AUROC, Tier, and r)

83.7MMLU AUROCClaude Sonnet 4.6
48.65257.75166.8575.949Apr 20, 2026

Evaluation results

MethodLinks
Claude Sonnet 4.62026.0483.7—0.25
Claude Opus 4.62026.0483.6—0.217
Gemini 2.5 Pro2026.0480.3—0.558
Gemini 3.1 Pro2026.0478.6—0.294
DeepSeek-R12026.0476.8—0.204
Claude Haiku 4.52026.0476—0.283
Gemini 2.5 Flash2026.0472.4—0.246
Qwen 80B Think2026.0471.8—0.265
GPT-5.42026.0471.5—0.19
GPT-5.4 mini2026.0469.9—0.235
Gemini 3 Flash2026.0469.5—0.201
GLM-52026.0467.9—0.135
Qwen 235B2026.0464.7—0.221
Qwen 80B Instruct2026.0461.3—0.141
Gemma 3 27B2026.0458.8—0.147
GPT-5.4 nano2026.0455.4—0.066
Gemma 3 12B2026.0453.3—0.106
Gemma 3 1B2026.0450—0