Qwen3-235B-A22B-Instruct Jul 2025
Qwen3-235B-A22B-Instruct places in the top three on 3 benchmarks.
46
Benchmarks with a score
0
Ranked first, out of five or more models
3
In the top three
Qwen Jul 2025 Qwen3-235B-A22B-Instruct
Jul 2025 snapshot Open weights
138.1
#132 of 256
Epoch Capabilities Index
- VITA-Bench 12.3% #16 /18
- ProfBench 45% #45 /72
- PRBench 38.3% #40 /48
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 | |
|---|---|---|---|---|
| Encyclo-K | 50.4% | #2 of 23 | Official leaderboard | |
| WritingBench | 80.3 | #6 of 45 | Official leaderboard | |
| LongBench v2 | 58.3% | #5 of 33 | Official leaderboard | |
| HUMAINE | 32 | #8 of 50 | Official leaderboard | |
| MMLU-Pro | 83% | #42 of 212 | Official leaderboard | |
| LMArena (Elo) | 1,422 | #103 of 335 | Official leaderboard | |
| HELM Arabic | 79.5 | #9 of 28 | Official leaderboard | |
| HELM Arabic Enterprise | 69.6 | #8 of 23 | Official leaderboard | |
| IFEval | 83.5% | #29 of 60 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 57.4 | #72 of 119 | Official leaderboard | |
| Fiction.LiveBench | 52.9% | #38 of 57 | Fiction.live leaderboard, via Epoch AI | |
| Judgemark v2.1 | 59.7 | #31 of 42 | Official leaderboard | |
| RULERv2 | 83.7% | #3 of 3 | Official leaderboard |
Math and science
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ArtifactsBench | 50.6% | #10 of 20 | Official leaderboard | |
| AutoCodeBench | 43.1% | #19 of 34 | Official leaderboard | |
| BigCodeArena | 1,004 | #8 of 13 | Official leaderboard | |
| WeirdML | 38.7% | #84 of 112 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 29 | #44 of 188 | Official leaderboard | |
| Galileo Agent Leaderboard v2 | 53% | #6 of 21 | Official leaderboard | |
| BFCL V4 | 52.1% | #19 of 65 | Official leaderboard | |
| OdysseyArena AI Trading | 11.3% | #3 of 8 | Official leaderboard | |
| Elimination Game | 4 | #24 of 56 | Official leaderboard | |
| Step Race | 2 | #27 of 59 | Official leaderboard | |
| OdysseyArena Energy Dispatch | 0% | #4 of 8 | Official leaderboard | |
| OdysseyArena Turn on Lights | 15% | #4 of 8 | Official leaderboard | |
| LLM Chess | -140 | #87 of 123 | Official leaderboard | |
| LisanBench | 133.3 | #106 of 133 | Official leaderboard | |
| OdysseyArena Repo System | 15.8% | #7 of 8 | Official leaderboard | |
| VITA-Bench | 12.3% | #16 of 18 | Official leaderboard | |
| MCP-Universe | 18.2% | #28 of 30 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| StockBench | 2.00 | #2 of 13 | Official leaderboard | |
| ProfBench | 45% | #45 of 72 | Official leaderboard | |
| PRBench | 38.3% | #40 of 48 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Anthropic Red Team | 99.9% | #5 of 75 | Official leaderboard | |
| LiveSecBench | 69.2 | #7 of 34 | Official leaderboard | |
| SimpleSafetyTests | 100% | #17 of 75 | Official leaderboard | |
| AIR-Bench 2024 | 79% | #25 of 82 | Official leaderboard | |
| HELM Safety | 94.8 | #23 of 75 | Official leaderboard | |
| SUPERChem | 26.4% | #5 of 6 | Official leaderboard | |
| Lakera B3 (Backbone Breaker Benchmark) | 44.9% | #30 of 35 | Official leaderboard | |
| Spiral-Bench v1.2 | 20.4 | #19 of 21 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Capabilities | 79.8 | #6 of 60 | Official leaderboard |
Qwen's models over time
Qwen's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)