Alibaba Cloud / Qwen Team

Qwen3 30B A3B

Zero-eval
#3Arena Hard

by Alibaba Cloud / Qwen Team

+
+
+
+
About

Qwen3 30B was created with a mixture-of-experts architecture featuring 30 billion total parameters and 3 billion active parameters. Built to provide strong capabilities with efficient inference through sparse activation, it balances performance with practical computational requirements.

+
+
+
+
Pricing Range
Input (per 1M)$0.10 -$0.89
Output (per 1M)$0.30 -$0.89
Providers3
+
+
+
+
Timeline
AnnouncedApr 29, 2025
ReleasedApr 29, 2025
+
+
+
+
Specifications
Training Tokens36.0T
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown

Overall Performance

8 benchmarks
Average Score
73.3%
Best Score
91.0%
High Performers (80%+)
2

Performance Metrics

Max Context Window
256.0K
Avg Throughput
97.9 tok/s
Avg Latency
1ms
+
+
+
+
All Benchmark Results for Qwen3 30B A3B
Complete list of benchmark scores with detailed information
Arena Hard
text
0.91
91.0%
Self-reported
AIME 2024
text
0.80
80.4%
Self-reported
LiveBench
text
0.74
74.3%
Self-reported
Multi-IF
text
0.72
72.2%
Self-reported
AIME 2025
text
0.71
70.9%
Self-reported
BFCL
text
0.69
69.1%
Self-reported
GPQA
text
0.66
65.8%
Self-reported
LiveCodeBench
text
0.63
62.6%
Self-reported
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+