Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Qwen3.5 397B
68
2/8 categoriesQwen3.6 Plus
69
Winner · 5/8 categoriesQwen3.5 397B· Qwen3.6 Plus
Pick Qwen3.6 Plus if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if reasoning is the priority or you would rather avoid the extra latency and token burn of a reasoning model.
Qwen3.6 Plus finishes one point ahead overall, 69 to 68. That is enough to call, but not enough to treat as a blowout. This matchup comes down to a few meaningful edges rather than one model dominating the board.
Qwen3.6 Plus's sharpest advantage is in coding, where it averages 64.9 against 52.2. The single biggest benchmark swing on the page is QwenWebBench, 1162 to 1502. Qwen3.5 397B does hit back in reasoning, so the answer changes if that is the part of the workload you care about most.
Qwen3.6 Plus is the reasoning model in the pair, while Qwen3.5 397B is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Qwen3.6 Plus gives you the larger context window at 1M, compared with 128K for Qwen3.5 397B.
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus |
|---|---|---|
| AgenticQwen3.6 Plus wins | ||
| Terminal-Bench 2.0 | 52.5% | 61.6% |
| BrowseComp | 62% | — |
| OSWorld-Verified | 62.2% | 62.5% |
| Claw-Eval | 48.1% | 58.7% |
| QwenClawBench | 51.8% | 57.2% |
| QwenWebBench | 1162 | 1502 |
| TAU3-Bench | 68.4% | 70.7% |
| VITA-Bench | 43.7% | 44.3% |
| DeepPlanning | 37.6% | 41.5% |
| Toolathlon | 36.3% | 39.8% |
| MCP Atlas | 46.1% | 48.2% |
| MCP-Tasks | 74.2% | 74.1% |
| WideResearch | 74.0% | 74.3% |
| CodingQwen3.6 Plus wins | ||
| HumanEval | 75% | — |
| SWE-bench Verified | 76.2% | 78.8% |
| LiveCodeBench | 39% | — |
| LiveCodeBench v6 | 83.6% | 87.1% |
| SWE-bench Pro | 50.9% | 56.6% |
| SWE Multilingual | 69.3% | 73.8% |
| NL2Repo | 32.2% | 37.9% |
| Multimodal & GroundedQwen3.6 Plus wins | ||
| MMMU | 85.0% | 86.0% |
| MMMU-Pro | 79% | 78.8% |
| OfficeQA Pro | 68% | — |
| RealWorldQA | 83.9% | 85.4% |
| OmniDocBench 1.5 | 90.8% | 91.2% |
| Video-MME (with subtitle) | 87.5% | 87.8% |
| Video-MME (w/o subtitle) | 84.2% | 84.2% |
| MathVision | 88.6% | 88.0% |
| We-Math | 87.9% | 89.0% |
| DynaMath | 86.3% | 88.0% |
| MStar | 83.8% | 83.3% |
| SimpleVQA | 67.1% | 67.3% |
| ChatCVQA | 80.8% | 81.5% |
| MMLongBench-Doc | 61.5% | 62.0% |
| CC-OCR | 82.0% | 83.4% |
| AI2D_TEST | 93.9% | 94.4% |
| CountBench | 97.2% | 97.6% |
| RefCOCO (avg) | 92.3% | 93.5% |
| ODINW13 | 47.0% | 51.8% |
| ERQA | 67.5% | 65.7% |
| VideoMMMU | 84.7% | 84.0% |
| MLVU (M-Avg) | 86.7% | 86.7% |
| ScreenSpot Pro | 65.6% | 68.2% |
| ReasoningQwen3.5 397B wins | ||
| MuSR | 78% | — |
| BBH | 82% | — |
| LongBench v2 | 63.2% | 62% |
| MRCRv2 | 71% | — |
| AI-Needle | 68.7% | 68.3% |
| KnowledgeQwen3.5 397B wins | ||
| MMLU | 83% | — |
| GPQA | 88.4% | 90.4% |
| SuperGPQA | 70.4% | 71.6% |
| MMLU-Pro | 87.8% | 88.5% |
| MMLU-Redux | 94.9% | 94.5% |
| C-Eval | 93% | 93.3% |
| HLE | 28.7% | 28.8% |
| FrontierScience | 71% | — |
| SimpleQA | 80% | — |
| Instruction FollowingQwen3.6 Plus wins | ||
| IFEval | 92.6% | 94.3% |
| IFBench | 76.5% | 74.2% |
| MultilingualQwen3.6 Plus wins | ||
| MGSM | 82% | — |
| MMLU-ProX | 84.7% | 84.7% |
| NOVA-63 | 59.1% | 57.9% |
| INCLUDE | 85.6% | 85.1% |
| PolyMath | 73.3% | 77.4% |
| VWT2k-lite | 78.9% | 84.3% |
| MAXIFE | 88.2% | 88.2% |
| Mathematics | ||
| AIME 2023 | 83% | — |
| AIME 2024 | 85% | — |
| AIME 2025 | 84% | — |
| AIME26 | 93.3% | 95.3% |
| HMMT Feb 2023 | 79% | — |
| HMMT Feb 2024 | 81% | — |
| HMMT Feb 2025 | 80% | — |
| HMMT Feb 2025 | 94.8% | 96.7% |
| HMMT Nov 2025 | 92.7% | 94.6% |
| HMMT Feb 2026 | 87.9% | 87.8% |
| MMAnswerBench | 80.9% | 83.8% |
| BRUMO 2025 | 82% | — |
| MATH-500 | 81% | — |
Qwen3.6 Plus is ahead overall, 69 to 68. The biggest single separator in this matchup is QwenWebBench, where the scores are 1162 and 1502.
Qwen3.5 397B has the edge for knowledge tasks in this comparison, averaging 68.2 versus 66. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Qwen3.6 Plus has the edge for coding in this comparison, averaging 64.9 versus 52.2. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.
Qwen3.5 397B has the edge for reasoning in this comparison, averaging 69.7 versus 62. Inside this category, LongBench v2 is the benchmark that creates the most daylight between them.
Qwen3.6 Plus has the edge for agentic tasks in this comparison, averaging 62 versus 58.3. Inside this category, QwenWebBench is the benchmark that creates the most daylight between them.
Qwen3.6 Plus has the edge for multimodal and grounded tasks in this comparison, averaging 78.8 versus 74.1. Inside this category, ODINW13 is the benchmark that creates the most daylight between them.
Qwen3.6 Plus has the edge for instruction following in this comparison, averaging 94.3 versus 92.6. Inside this category, IFBench is the benchmark that creates the most daylight between them.
Qwen3.6 Plus has the edge for multilingual tasks in this comparison, averaging 84.7 versus 83.8. Inside this category, VWT2k-lite is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.