Qwen3-Next-80B-A3B-Thinking
Zero-eval
#2AIME25
#2LiveBench 20241125
#2LiveCodeBench v6
+9 more
by Alibaba Cloud / Qwen Team
About
Qwen3-Next-80B-A3B-Thinking is a language model developed by Alibaba Cloud / Qwen Team. It achieves strong performance with an average score of 67.9% across 23 benchmarks. It excels particularly in MMLU-Redux (92.5%), IFEval (88.9%), AIME25 (87.8%). It supports a 131K token context window for handling large documents. The model is available through 1 API provider. It's licensed for commercial use, making it suitable for enterprise applications. Released in 2025, it represents Alibaba Cloud / Qwen Team's latest advancement in AI technology.
Pricing Range
Input (per 1M)$0.15 -$0.15
Output (per 1M)$1.50 -$1.50
Providers1
Timeline
AnnouncedJan 10, 2025
ReleasedJan 10, 2025
Specifications
Training Tokens15.0T
License & Family
License
Apache 2.0
Benchmark Performance Overview
Performance metrics and category breakdown
Overall Performance
23 benchmarks
Average Score
67.9%
Best Score
92.5%
High Performers (80%+)
5Performance Metrics
Max Context Window
131.1KTop Categories
factuality
92.5%
general
77.6%
roleplay
76.6%
reasoning
60.8%
agents
58.2%
Additional Information
Content coming soon...
All Benchmark Results for Qwen3-Next-80B-A3B-Thinking
Complete list of benchmark scores with detailed information
| MMLU-Redux | factuality | text | 0.93 | 92.5% | Self-reported |
| IFEval | code | text | 0.89 | 88.9% | Self-reported |
| AIME25 | general | text | 0.88 | 87.8% | Self-reported |
| WritingBench | general | text | 0.85 | 84.6% | Self-reported |
| MMLU-Pro | general | text | 0.83 | 82.7% | Self-reported |
| Include | general | text | 0.79 | 78.9% | Self-reported |
| MMLU-ProX | general | text | 0.79 | 78.7% | Self-reported |
| Multi-IF | general | text | 0.78 | 77.8% | Self-reported |
| GPQA | general | text | 0.77 | 77.2% | Self-reported |
| LiveBench 20241125 | roleplay | text | 0.77 | 76.6% | Self-reported |
Showing 1 to 10 of 23 benchmarks