Atlas

Models

← All models

QwQ-32B-Preview

Alibaba · 2024-11-28 · 32.8B parameters

QwQ-32B-Preview is Alibaba's November 2024 preview reasoning checkpoint for mathematical, scientific, and coding tasks. It is distinct from the final QwQ-32B checkpoint released in March 2025.

Benchmark scores

BenchmarkScore
AIME 202533.3
Arena Text — Business, Management & Financial Ops (No Style Control)1098
Arena Text — Chinese (No Style Control)1217
Arena Text — Creative Writing (No Style Control)1134
Arena Text — English (No Style Control)1185
Arena Text — Entertainment, Sports & Media (No Style Control)1115
Arena Text — Exclude Ties (No Style Control)1023
Arena Text — Exclude Ties (Style Controlled)1012
Arena Text — Expert (No Style Control)1134
Arena Text — Hard Prompts (No Style Control)1163
Arena Text — Hard Prompts English (No Style Control)1173
Arena Text — Instruction Following (No Style Control)1156
Arena Text — Legal & Government (No Style Control)1143
Arena Text — Life, Physical & Social Science (No Style Control)1154
Arena Text — Longer Query (No Style Control)1167
Arena Text — Multi-turn (No Style Control)1141
Arena Text — Non-English (No Style Control)1129
Arena Text — Non-English (Style Controlled)1126
Arena Text — Occupational Mathematical (No Style Control)1222
Arena Text — Russian (No Style Control)1114
Arena Text — Software & IT Services (No Style Control)1155
Arena Text — Writing, Literature & Language (No Style Control)1139
Artificial Analysis Intelligence Index9.2
BBH (Open LLM Leaderboard v2)66.9
BigCodeBench Complete54.4
BigCodeBench Instruct44.6
BigCodeBench-Hard Complete28.4
BigCodeBench-Hard Instruct25.0
BRIDGE Medical (chain-of-thought)23.3
BRIDGE Medical (few-shot)48.4
Loading Atlas data…