Qwen3-Max: Just Scale it (blog) Sep 24, 2025
Qwen3-Max
Qwen3-Max ranks first on 2 of the 86 benchmarks with five or more models scored.
86
Benchmarks with a score
2
Ranked first, out of five or more models
6
In the top three
Qwen Sep 2025 Qwen3-Max
Closed weights Flagship
142.6
#110 of 256
Epoch Capabilities Index
Self-Reported Index 150.3 #90 of 404
- IFBench 70.9% #41 /80
- AA-LCR (Artificial Analysis Long Context Reasoning) 68.7% #10 /38
- BFCL V4 67.7% #11 /30
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 | |
|---|---|---|---|---|
| C-Eval | 93.7% | #1 of 46 | Lab model card | |
| Arena-Hard v2.0 | 90.2% | #1 of 24 | Lab model card | |
| MMLU-Pro | 85.7% | #21 of 167 | Lab model card | |
| WritingBench | 80.3 | #7 of 45 | Official leaderboard | |
| BullshitBench V2 | 63% | #32 of 129 | Official leaderboard | |
| LMArena (Elo) | 1,434 | #85 of 335 | Official leaderboard | |
| AA-LCR (Artificial Analysis Long Context Reasoning) | 68.7% | #10 of 38 | Lab model card | |
| Humanity's Last Exam | 49.8% | #43 of 154 | Lab model card | |
| MultiChallenge | 63.3% | #13 of 42 | Lab model card | |
| KINA | 35.9% | #13 of 37 | Official leaderboard | |
| SimpleQA Verified | 48.8% | #34 of 94 | Epoch AI run | |
| MMLU-Redux | 92.8% | #17 of 45 | Lab model card | |
| Vectara Hallucination Leaderboard (HHEM) | 3.80% | #61 of 148 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 66.2 | #51 of 119 | Official leaderboard | |
| Fiction.LiveBench | 66.7% | #27 of 57 | Fiction.live leaderboard, via Epoch AI | |
| IFBench | 70.9% | #41 of 80 | Lab model card | |
| Phare | 65.9% | #33 of 62 | Official leaderboard | |
| Lech Mazur Creative Writing | -0.73 | #32 of 50 | Official leaderboard | |
| HLE-Verified (text-only Full) | 38.2% | #4 of 6 | Official leaderboard | |
| AMemGym | 41.4% | #8 of 11 | Official leaderboard | |
| DesignArena ASCII Art | 1,122 | #59 of 77 | Official leaderboard | |
| CL-bench | 14.5% | #20 of 24 | Official leaderboard | |
| Hemingway-bench | 948 | #38 of 42 | Official leaderboard | |
| Extended NYT Connections | 11.8% | #96 of 103 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 100% | #3 of 143 | Lab model card | |
| HMMT 2025 | 100% | #2 of 54 | Lab model card | |
| MATH Level 5 | 97.1% | #6 of 94 | Epoch AI run | |
| HMMT Nov 2025 | 94.7% | #3 of 28 | Lab model card | |
| SGI-Bench | 32% | #3 of 15 | Official leaderboard | |
| DTBench | 87.7% | #60 of 180 | Official leaderboard | |
| General365 | 57.2% | #9 of 23 | Official leaderboard | |
| IMO-AnswerBench | 83.9% | #16 of 37 | Lab model card | |
| OTIS Mock AIME 2024-2025 | 73.3% | #103 of 212 | Epoch AI run | |
| GPQA Diamond | 72.6% | #114 of 229 | Epoch AI run | |
| AMO-Bench | 28.8% | #11 of 21 | Official leaderboard | |
| PHYBench | 22.1% | #15 of 28 | Official leaderboard | |
| Echo General AI Prediction Leaderboard | 966 | #15 of 17 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 18.9% | #77 of 86 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 85.9% | #18 of 90 | Lab model card | |
| SWE-bench Verified | 75.3% | #43 of 155 | Lab model card | |
| Spider 2.0 | 45.1% | #10 of 23 | Official leaderboard | |
| BigCodeArena | 1,023 | #6 of 13 | Official leaderboard | |
| IOI (Vals) | 15.7% | #27 of 55 | Official leaderboard | |
| DesignArena Game Dev | 1,084 | #106 of 143 | Official leaderboard | |
| DesignArena Website | 1,112 | #112 of 151 | Official leaderboard | |
| Vibe Code Bench v1 | 3.51% | #13 of 17 | Official leaderboard | |
| DesignArena Code | 1,100 | #111 of 143 | Official leaderboard | |
| DesignArena 3D Design | 1,057 | #106 of 136 | Official leaderboard | |
| DesignArena UI Component | 1,069 | #111 of 139 | Official leaderboard | |
| DesignArena Data Visualization | 1,098 | #102 of 118 | Official leaderboard | |
| DesignArena SVG | 1,002 | #93 of 103 | Official leaderboard | |
| Vibe Code Bench | 3.51% | #110 of 120 | Official leaderboard | |
| PostTrainBench | 7.42% | #22 of 23 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 31 | #33 of 188 | Official leaderboard | |
| Elimination Game | 4 | #20 of 56 | Official leaderboard | |
| BFCL V4 | 67.7% | #11 of 30 | Lab model card | |
| tau2-bench | 82.1% | #29 of 79 | Lab model card | |
| DeepPlanning | 29.7% | #7 of 15 | Official leaderboard | |
| BalatroBench | 7.27 | #9 of 16 | Official leaderboard | |
| Step Race | 2 | #37 of 59 | Official leaderboard | |
| MCPMark | 17.7% | #18 of 28 | Official leaderboard | |
| ClawProBench | 55.8% | #37 of 49 | Official leaderboard | |
| AutoResearchBench | 3.24% | #10 of 13 | Official leaderboard | |
| VITA-Bench | 14.3% | #14 of 18 | Official leaderboard | |
| CompileBench | 66.7% | #16 of 20 | Official leaderboard | |
| VitaBench 2.0 | 32.4% | #14 of 17 | Official leaderboard | |
| Agent Poker Bench | 995 | #15 of 17 | Official leaderboard | |
| Mystery Game Puzzles | 5.00% | #82 of 84 | Epoch AI run | |
| RuneBench | 1.40 | #75 of 76 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CorpFin v2 | 68% | #12 of 119 | Official leaderboard | |
| Vals Finance Agent v1 | 43.2% | #15 of 47 | Official leaderboard | |
| TaxEval v2 | 74% | #42 of 129 | Official leaderboard | |
| PLawBench | 64.8% | #4 of 11 | Official leaderboard | |
| LegalBench | 81.9% | #76 of 138 | Official leaderboard | |
| CaseLaw v2 | 55% | #35 of 53 | Official leaderboard | |
| Vals Finance Agent | 44.3% | #38 of 49 | Official leaderboard | |
| MedScribe | 72.7% | #78 of 97 | Official leaderboard | |
| MedCode | 31.4% | #87 of 96 | Official leaderboard | |
| Vending-Bench 2 | $72 | #61 of 67 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 3.60% | #50 of 51 | Official leaderboard | |
| Finch (FinWorkBench) | 14.5% | #6 of 6 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 990 | #68 of 75 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 83.3 | #49 of 185 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 48.6 | #74 of 151 | Official leaderboard | |
| ACCoRD | 62.8 | #81 of 152 | Official leaderboard | |
| LMCA | 33.6% | #84 of 147 | Official leaderboard |
Its lab's numbers put it at 150.3, 7.7 points above its ECI
The Self-Reported Index fits one score to the 19 capability results Qwen reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 147.9 to 152.7. How we compute it.
- AIME 2025 100
- HMMT 2025 100
- HMMT Nov 2025 94.7
- C-Eval 93.7
- MMLU-Redux 92.8
- Arena-Hard v2.0 90.2
- GPQA Diamond 87.4
- LiveCodeBench v6 85.9
- MMLU-Pro 85.7
- IMO-AnswerBench 83.9
- tau2-bench 82.1
- SWE-bench Verified 75.3
- IFBench 70.9
- AA-LCR (Artificial Analysis Long Context Reasoning) 68.7
- BFCL V4 67.7
- MultiChallenge 63.3
- Humanity's Last Exam 49.8
- VITA-Bench 40.9
- DeepPlanning 28.7
Its lab's numbers next to 5 leaderboards
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap | What differs |
|---|---|---|---|---|
| VITA-Bench | 40.9% Vita Bench | 14.3% Leaderboard, non-thinking | +27 pts | The lab reports Qwen3-Max-Thinking, while the board evaluates Qwen3-Max without thinking. |
| GPQA Diamond | 87.4% GPQA | 72.6% Epoch AI run | +15 pts | |
| DeepPlanning | 28.7% Deep Planning | 29.7% Leaderboard, v1.1; thinking | −1.0 pts | |
| tau2-bench | 82.1% Tau² Bench | 82.2% Leaderboard | Same | |
| MMLU-Pro | 85.7% MMLU-Pro | 85.7% Leaderboard | Same |
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)
Its launch posts
The benchmarks the lab chose to headline, with the scores it gave.
Pushing Qwen3-Max-Thinking Beyond its Limits (blog) Jan 23, 2026
- MMLU-Pro 85.7
- MMLU-Redux 92.8
- C-Eval 93.7
- GPQA 87.4
- Best of HLE / HLE (w/ tools) 49.8
- LiveCodeBench v6 85.9
- HMMT Feb 25 98
- HMMT Nov 25 94.7
- IMOAnswerBench 83.9
- SWE Verified 75.3
- IFBench 70.9
- MultiChallenge 63.3
- Arena-Hard v2 90.2
- Tau² Bench 82.1
- BFCL-V4 67.7
- Vita Bench 40.9
- Deep Planning 28.7
- AA-LCR 68.7