- SuperGPQA Graduate-Level Knowledge 73.9
- AA-Omniscience Index Knowledge Reliability and Hallucination 10
- GDPval-AA Real-World Valuable Task (51.0, scale not stated) 51
- QwenChineseBench Chinese Real-World Knowledge 84
- QwenClawBench Real-World Agent 59
- SkillsBench Agent Skills 55.6
- ToolcallFormatIFBench Real-World Toolcall Following 86.1
- QwenWebBench (Elo Rating) Artifacts 1532
- SciCode Research Coding 47
- NL2Repo Long-Horizon Coding 42.9
- Terminal-Bench 2.0 (Terminus-2) Agentic Terminal Coding 65.4
- SWE-bench Pro Agentic Coding 57.3
Qwen 3.6 Max (Preview)
Qwen 3.6 Max (Preview) ranks first on 2 of the 36 benchmarks with five or more models scored.
38
Benchmarks with a score
2
Ranked first, out of five or more models
3
In the top three
Qwen Apr 2026 Qwen 3.6 Max (Preview)
149.8
#72 of 256
Epoch Capabilities Index
Self-Reported Index 152.3 #80 of 404
- SWE-bench Verified 76.7% #11 /32
- SciCode 47% #17 /49
- NL2Repo-Bench 42.9% #19 /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 | |
|---|---|---|---|---|
| SuperGPQA | 73.9% | #1 of 35 | Lab model card | |
| LMArena (Elo) | 1,460 | #50 of 335 | Official leaderboard | |
| SimpleQA Verified | 52% | #28 of 94 | Epoch AI run | |
| AA-Omniscience | 10.0 | #4 of 10 | Lab model card | |
| Extended NYT Connections | 74.1% | #56 of 103 | Official leaderboard | |
| Lech Mazur Creative Writing | -0.98 | #34 of 50 | Official leaderboard | |
| Judgemark v4 | 45.1 | #29 of 40 | Official leaderboard | |
| QwenChineseBench | 84% | #1 of 1 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SimpleBench | 63% | #32 of 106 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 91.1% | #65 of 212 | Epoch AI run | |
| GPQA Diamond | 87.4% | #71 of 229 | Epoch AI run | |
| DTBench | 87.2% | #66 of 180 | Official leaderboard | |
| FrontierMath | 23.1% | #22 of 55 | Official leaderboard | |
| FrontierMath Tier 4 | 4.17% | #35 of 55 | Epoch AI run |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| QwenWebBench | 1,532 | #1 of 4 | Lab model card | |
| SWE-bench Verified | 76.7% | #11 of 32 | Epoch AI run | |
| SciCode | 47% | #17 of 49 | Lab model card | |
| LiveSQLBench | 26.5% | #15 of 36 | Official leaderboard | |
| WebDev Arena | 1,479 | #48 of 105 | arena.ai, via Epoch AI | |
| SWE-Bench Pro | 57.3% | #39 of 81 | Lab model card | |
| Code Arena | 1,482 | #64 of 128 | Official leaderboard | |
| NL2Repo-Bench | 42.9% | #19 of 30 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 31 | #32 of 188 | Official leaderboard | |
| ToolcallFormatIFBench | 86.1% | #1 of 5 | Official leaderboard | |
| Terminal-Bench 2.0 | 65.4% | #15 of 63 | Lab model card | |
| SkillsBench | 55.6% | #7 of 14 | Lab model card | |
| QwenClawBench | 59% | #3 of 6 | Lab model card | |
| Mystery Game Puzzles | 19% | #52 of 84 | Epoch AI run | |
| ClawProBench | 57.4% | #31 of 49 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CorpFin v2 | 66.5% | #19 of 119 | Official leaderboard | |
| Vals Finance Agent | 52.8% | #22 of 49 | Official leaderboard | |
| CHI-Bench | 16.4% | #13 of 23 | Official leaderboard | |
| Vending-Bench 2 | $4,254 | #40 of 67 | Official leaderboard | |
| CaseLaw v2 | 47.9% | #51 of 53 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 87.0 | #39 of 185 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ACCoRD | 70.8 | #50 of 152 | Official leaderboard | |
| LMCA | 42.5% | #50 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 55.4 | #54 of 151 | Official leaderboard |
Its lab's numbers put it at 152.3, 2.5 points above its ECI
The Self-Reported Index fits one score to the 6 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 149.1 to 155.5. How we compute it.
- SuperGPQA 73.9
- Terminal-Bench 2.0 65.4
- SWE-Bench Pro 57.3
- SkillsBench 55.6
- SciCode 47
- NL2Repo-Bench 42.9
Its lab's number next to a leaderboard
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 |
|---|---|---|---|
| ToolcallFormatIFBench | 86.1% ToolcallFormatIFBench Real-World Toolcall Following | 86.1% 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 post
The benchmarks the lab chose to headline, with the scores it gave.