Alibaba Cloud / Qwen Team

Qwen3 32B

Zero-eval
#1MultiLF
#2Arena Hard

by Alibaba Cloud / Qwen Team

+
+
+
+
About

Qwen3 32B was developed as a dense 32-billion-parameter model in the Qwen3 family, designed to provide strong language understanding without mixture-of-experts complexity. Built for applications requiring straightforward deployment and reliable performance, it serves as a capable mid-to-large-scale foundation model.

+
+
+
+
Pricing Range
Input (per 1M)$0.10 -$0.40
Output (per 1M)$0.30 -$0.80
Providers3
+
+
+
+
Timeline
AnnouncedApr 29, 2025
ReleasedApr 29, 2025
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown

Overall Performance

9 benchmarks
Average Score
72.0%
Best Score
93.8%
High Performers (80%+)
2

Performance Metrics

Max Context Window
256.0K
Avg Throughput
129.0 tok/s
Avg Latency
1ms
+
+
+
+
All Benchmark Results for Qwen3 32B
Complete list of benchmark scores with detailed information
Arena Hard
text
0.94
93.8%
Self-reported
AIME 2024
text
0.81
81.4%
Self-reported
LiveBench
text
0.75
74.9%
Self-reported
MultiLF
text
0.73
73.0%
Self-reported
AIME 2025
text
0.73
72.9%
Self-reported
BFCL
text
0.70
70.3%
Self-reported
CodeForces
text
0.66
65.9%
Self-reported
LiveCodeBench
text
0.66
65.7%
Self-reported
Aider
text
0.50
50.2%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+