- MMLU-Pro 82.7
- MMLU-Redux 92.5
- GPQA 77.2
- SuperGPQA 60.8
- AIME25 87.8
- HMMT25 73.9
- LiveBench 241125 76.6
- LiveCodeBench v6 (25.02-25.05) 68.7
- CFEval 2071
- OJBench 29.7
- IFEval 88.9
- Arena-Hard v2* (win rate judged by GPT-4.1) 62.3
- WritingBench 84.6
- BFCL-v3 72
- Mean of TAU1-Retail 69.6 / TAU1-Airline 49.0 59.3
- Mean of TAU2-Retail 67.8 / TAU2-Airline 60.5 / TAU2-Telecom 43.9 57.4
- MultiIF 77.8
- MMLU-ProX 78.7
- INCLUDE 78.9
- PolyMATH 56.3
Qwen3-Next-80B-A3B
Qwen3-Next-80B-A3B places in the top three on WritingBench.
37
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Qwen Sep 2025 Qwen3-Next-80B-A3B
Open weights Mid-size
No Epoch Capabilities Index score yet
- SuperGPQA 60.8% #17 /26
- AMO-Bench 34.8% #8 /21
- Vectara Hallucination Leaderboard (HHEM) 3.70% #96 /148
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| WritingBench | 84.6 | #3 of 14 | Model card | |
| Multi-IF | 77.8% | #8 of 28 | Model card | |
| IFEval | 88.9% | #40 of 126 | Model card | |
| MMLU-Pro | 82.7% | #50 of 151 | Model card | |
| MMLU-Redux | 92.5% | #17 of 38 | Model card | |
| LMArena (Elo) | 1,369 | #159 of 333 | Official leaderboard | |
| INCLUDE | 78.9% | #12 of 25 | Model card | |
| Vectara Hallucination Leaderboard (2025 dataset) | 9.30% | #50 of 104 | Official leaderboard | |
| MMLU-ProX | 78.7% | #11 of 17 | Model card | |
| Vectara Hallucination Leaderboard (HHEM) | 3.70% | #96 of 148 | Official leaderboard | |
| SuperGPQA | 60.8% | #17 of 26 | Model card | |
| Arena-Hard v2.0 | 62.3% | #16 of 24 | Model card |
Math and science
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Codeforces rating | 2,071 | #22 of 41 | Model card | |
| LiveCodeBench v6 | 68.7% | #51 of 76 | Model card | |
| OJBench | 29.7% | #9 of 12 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Berkeley Function Calling Leaderboard (BFCL) | 72% | #14 of 47 | Model card | |
| SnakeBench | 26 | #58 of 183 | Official leaderboard | |
| tau-bench | 59.3% | #12 of 23 | Model card | |
| LLM Chess | -128 | #80 of 124 | Official leaderboard | |
| Pencil Puzzle Bench | 0% | #38 of 53 | Official leaderboard | |
| tau2-bench | 57.4% | #62 of 69 | Model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Physical-Engineering's Last Exam | 0.04 | #17 of 18 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Anthropic Red Team | 99.9% | #4 of 75 | Official leaderboard | |
| AIR-Bench 2024 | 86.7% | #8 of 82 | Official leaderboard | |
| HELM Safety | 94.8 | #24 of 75 | Official leaderboard | |
| SimpleSafetyTests | 99.5% | #27 of 75 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 76.6% | #6 of 27 | Model card | |
| HELM Capabilities | 70.0 | #21 of 60 | Official leaderboard | |
| Political Science LLM Benchmark | 65.1% | #16 of 29 | Official leaderboard | |
| ACCoRD | 59.6 | #89 of 150 | Official leaderboard | |
| Conceptual Reasoning Index | 36.7 | #103 of 149 | Official leaderboard |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.