Alibaba Cloud / Qwen Team

QwQ-32B

Zero-eval

by Alibaba Cloud / Qwen Team

+
+
+
+
About

QwQ 32B was developed as a reasoning-focused model, designed to emphasize analytical thinking and problem-solving capabilities. Built with 32 billion parameters optimized for step-by-step reasoning, it demonstrates Qwen's exploration into models that prioritize deliberate analytical processing.

+
+
+
+
Timeline
AnnouncedMar 5, 2025
ReleasedMar 5, 2025
Knowledge CutoffNov 28, 2024
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown

Overall Performance

7 benchmarks
Average Score
74.6%
Best Score
90.6%
High Performers (80%+)
2
+
+
+
+
All Benchmark Results for QwQ-32B
Complete list of benchmark scores with detailed information
MATH-500
text
0.91
90.6%
Self-reported
IFEval
text
0.84
83.9%
Self-reported
AIME 2024
text
0.80
79.5%
Self-reported
LiveBench
text
0.73
73.1%
Self-reported
BFCL
text
0.66
66.4%
Self-reported
GPQA
text
0.65
65.2%
Self-reported
LiveCodeBench
text
0.63
63.4%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+