Qwen3.6-35B-A3B vs Qwen3.5-9B vs Qwen3.5-35B-A3B
New ComparisonScores are fetched live from the current database on every visit — nothing is cached.
Winner
Qwen3.6-35B-A3B
Qwen3.6-35B-A3B wins this comparison with an overall score of 86 vs 67 for Qwen3.5-35B-A3B, leading 18 of 22 shared benchmarks.
Qwen3.6-35B-A3B
85.6
Qwen3.5-35B-A3B
66.7
Qwen3.5-9B
27.2
Based on 22 of 130 benchmarks scored by all 3 selected models.
Benchmark Scores Comparison
Qwen3.6-35B-A3B
Qwen3.5-9B
Qwen3.5-35B-A3B
Qwen3.6-35B-A3B
35.0B params
Score: 100.0
Qwen3.5-9B
Score: 15.0
Qwen3.5-35B-A3B
35.0B params
Score: 100.0
Coverage matrix
3 models × 130 benchmarks — hover a benchmark name for its category
Shading = score relative to the best score among the selected models:
Low
Mid
High
Missing (—)
The totals row shows how many of the 130 benchmarks each model has scores for.
| Benchmark | Qwen3.5-35B-A3B Qwen Team (Alibaba Cloud) | Qwen3.6-35B-A3B Qwen Team (Alibaba Cloud) | Qwen3.5-9B Qwen Team (Alibaba Cloud) |
|---|---|---|---|
| HMMT Feb 25 | 89 | 90.7 | — |
| DynaMath | 85 | 82.8 | — |
| GPQA Diamond | 84.2 | 86 | — |
| MathVista (mini) | 86.2 | 86.4 | — |
| MMBench EN-DEV v1.1 | 91.5 | 92.8 | — |
| Nuscene | 14.6 | — | — |
| LiveCodeBench v6 | 74.6 | 80.4 | — |
| SWE-bench Pro | 44.6 | 49.5 | 33.8 |
| BFCL-V4 | 67.3 | 64.7 | 66.1 |
| RealWorldQA | 84.1 | 85.3 | — |
| EmbSpatialBench | 83.1 | 84.3 | — |
| AA-LCR | 58.5 | 69.3 | 63 |
| Hypersim | 13.1 | — | — |
| MathVision | 83.9 | — | — |
| AIME 26 | 91 | 92.7 | 88.2 |
| BrowseComp-zh | 69.5 | — | — |
| MAXIFE | 86.6 | — | — |
| AndroidWorld | 71.1 | — | — |
| AgentWorldBench Terminal | 46.13 | — | — |
| Toolathlon Verified | — | 41.7 | — |
| VITA-Bench | 29.1 | 35.6 | 15.6 |
| TAU3-Bench | 68.9 | 67.2 | 9.3 |
| ScreenSpot Pro | 68.6 | — | — |
| OJBench | 36 | — | — |
| AgentWorldBench Overall | 47.73 | — | — |
| AutomationBench (Public) | — | 24.9 | — |
| MMStar | 81.9 | 80.7 | — |
| GDPVal | — | 67.4 | — |
| OCRBench | 91 | — | — |
| Seal-0 | 41.4 | — | — |
| PMC-VQA | 62 | — | — |
| MMLongBench-Doc | 59.5 | — | — |
| SWE Atlas - TW | 9.8 | 13.3 | 4.4 |
| WildClawBench | — | 68.7 | — |
| AgentWorldBench Web | 47.1 | — | — |
| SWE Atlas - QnA | 13.2 | 15.5 | 9.2 |
| AI2D_TEST | 92.6 | 92.7 | — |
| LingoQA | 79.2 | — | — |
| ParseBench Text Formatting | — | 58.3 | — |
| NOVA-63 | 57.1 | — | — |
| NL2Repo | 20.5 | 29.4 | 16.2 |
| MMMU-Pro | 75.1 | 75.3 | — |
| INCLUDE | 79.7 | — | — |
| QwenClawBench | 47.7 | 52.6 | — |
| BabyVision | 38.4 | — | — |
| SLAKE | 78.7 | — | — |
| MMMLU | 85.2 | — | — |
| Claw-Eval Avg | 65.4 | 68.7 | 53.2 |
| VideoMMMU | 80.4 | 83.7 | — |
| HMMT Nov 25 | 89.2 | 89.1 | — |
| Multi-Challenge | 60 | 58.2 | — |
| PolyMATH | 64.4 | — | — |
| MMLU-ProX | 81 | — | — |
| MMLU-Pro | 85.3 | 85.2 | — |
| CodeForces | 2028 | — | — |
| SuperGPQA | 63.4 | 64.7 | — |
| OSWorld-Verified | 54.5 | — | — |
| TIR-Bench | 55.5 | — | — |
| Global PIQA | 86.6 | — | — |
| AgentWorldBench Search | 25.98 | — | — |
| IFEval | 91.9 | — | 91.5 |
| HLE (with tools) | 47.4 | 28.9 | — |
| LVBench | 71.4 | 71.4 | — |
| CharXiv (RQ) | 77.5 | 78 | — |
| GDP.pdf | — | 6.6 | — |
| BrowseComp | 61 | 62 | 8.3 |
| MedXpertQA-MM | 61.4 | — | — |
| MMLU-Redux | 93.3 | 93.3 | — |
| ParseBench Text Content | — | 90.7 | — |
| MCPMark | 27 | 37 | 13.4 |
| AA-Omniscience Index (public set) | — | -7.2 | — |
| SWE-bench Multilingual | 60.3 | 67.2 | 43.3 |
| MMVU | 72.3 | — | — |
| FullStackBench zh | 55 | — | — |
| MLVU | 85.6 | 86.2 | — |
| IFBench | 70.2 | 66.6 | 64.5 |
| FullStackBench en | 58.1 | — | — |
| SWE-bench Verified | 70 | 73.4 | 53.1 |
| IMOAnswerBench | 76.8 | 78.9 | 69.8 |
| Humanity's Last Exam | 22.4 | 21.4 | 14.3 |
| CountBench | 97.8 | 96.1 | — |
| ERQA | 64.8 | 61.8 | — |
| VideoMME (w/o sub.) | 82.5 | 82.5 | — |
| Terminal-Bench 2.0 | 40.5 | 51.5 | — |
| DeepPlanning | 22.8 | 25.9 | — |
| QwenWebBench | 978 | 1397 | — |
| TAU2-Bench | 81.2 | — | 79.1 |
| ODInW13 | 42.6 | 50.8 | — |
| VlmsAreBlind | 97 | 96.6 | — |
| Terminal-Bench 2.1 (Terminus-2) | 41.4 | 52.5 | 21.3 |
| DeepSWE | — | 0 | — |
| OmniDocBench 1.5 | 89.3 | 89.9 | — |
| HallusionBench | 67.9 | 69.8 | — |
| OCRBench | 89.4 | 90 | — |
| HMMT Feb 26 | 78.7 | 83.6 | 70.8 |
| Terminal-Bench Lite | — | 63.3 | — |
| AgentWorldBench OS | 56.27 | — | — |
| AgentWorldBench MCP | 57.87 | — | — |
| GPQA | — | — | 77.2 |
| AgentWorldBench SWE | 47.58 | — | — |
| CC-OCR | 80.7 | 81.9 | — |
| RefCOCO (avg) | 89.2 | 92 | — |
| Artificial Analysis Intelligence Index | — | 32 | — |
| Tool Decathlon | 28.7 | 26.9 | — |
| Frontier-Bench v0.1 | — | 1.4 | — |
| ParseBench Mean | — | 44.1 | — |
| ZEROBench_sub | 34.1 | 34.4 | — |
| C-Eval | 90.2 | 90 | — |
| WideSearch | 59.1 | 60.1 | — |
| SimpleVQA | 58.3 | 58.9 | — |
| VideoMME (w sub.) | 86.6 | 86.6 | — |
| Workspace Bench | — | — | 25.5 |
| SkillsBench Avg5 | 4.4 | 28.7 | — |
| Claw-Eval Pass^3 | 51 | 50 | — |
| WMT24++ (en→xx) | 76.3 | — | — |
| LongBench v2 | 59 | — | — |
| MCP-Atlas | 62.4 | 62.8 | 47.4 |
| SWE Atlas - RF | 10.2 | 11.4 | 4.3 |
| V-Star | 92.7 | 90.1 | — |
| AgentWorldBench Android | 53.18 | — | — |
| Terminal-Bench 2.1 (Claude Code) | 38.9 | 49.2 | 18.9 |
| Chatbot Korean (internal) | — | 74.5 | — |
| Gaokao 2026 | — | — | 135.5 |
| SUNRGBD | 33.4 | — | — |
| SciCode | — | 35.6 | 32.7 |
| RefSpatialBench | 63.5 | 64.3 | — |
| BrowseComp Plus | — | 34.8 | — |
| MMMU | 81.4 | 81.7 | — |
| MVBench | 74.8 | 74.6 | — |
| ZEROBench | 8 | — | — |
| Covered | 111/130 | 85/130 | 28/130 |
Scorecard
Overall 0–100 score — category strengths (50%), shared-benchmark wins (30%), coverage (20%)
| Model | Overall score | Shared benchmark wins | Coverage | Strongest at | Weakest at |
|---|---|---|---|---|---|
| Qwen3.6-35B-A3B Winner | 85.6 | 18 of 22 | 85 of 130 (65.0%) | Other | Instruction Following |
| Qwen3.5-35B-A3B | 66.7 | 4 of 22 | 111 of 130 (85.0%) | Instruction Following | Agentic Coding |
| Qwen3.5-9B | 27.2 | 0 of 22 | 28 of 130 (22.0%) | Long Context | Agentic Coding |
Category strengths
Where each model stands, per benchmark category — score relative to the best model, adjusted for benchmark coverage. This also ranks models when they share few benchmarks.
| Category | Qwen3.6-35B-A3B | Qwen3.5-35B-A3B | Qwen3.5-9B |
|---|---|---|---|
| General Agents 27 benchmarks | 93.3 | 91.2 | 43.4 |
| Agentic Coding 21 benchmarks | 99.7 | 66.3 | 26.4 |
| Vision & Language 19 benchmarks | 95.9 | 96.9 | — |
| STEM & Reasoning 11 benchmarks | 93.2 | 96.5 | 59.5 |
| Document Understanding 10 benchmarks | 100 | 96.3 | — |
| Spatial Intelligence 10 benchmarks | 94.5 | 81.7 | — |
| Video Understanding 7 benchmarks | 99.8 | 97.7 | — |
| Knowledge 5 benchmarks | 99.3 | 99.2 | — |
| Instruction Following 3 benchmarks | 77.4 | 100 | 79.6 |
| Long Context 3 benchmarks | 100 | 84.4 | 90.9 |
| Other 3 benchmarks | 100 | — | 90.2 |
Fully comparable benchmarks
Scored by all 3 selected models (22)
| Benchmark | Category | Qwen3.6-35B-A3B | Qwen3.5-9B | Qwen3.5-35B-A3B |
|---|---|---|---|---|
| SWE-bench Pro | coding_agent | 49.5 | 33.8 | 44.6 |
| BFCL-V4 | general_agent | 64.7 | 66.1 | 67.3 |
| AA-LCR | long_context | 69.3 | 63 | 58.5 |
| AIME 26 | stem_reasoning | 92.7 | 88.2 | 91 |
| VITA-Bench | general_agent | 35.6 | 15.6 | 29.1 |
| TAU3-Bench | general_agent | 67.2 | 9.3 | 68.9 |
| SWE Atlas - TW | coding_agent | 13.3 | 4.4 | 9.8 |
| SWE Atlas - QnA | coding_agent | 15.5 | 9.2 | 13.2 |
| NL2Repo | coding_agent | 29.4 | 16.2 | 20.5 |
| Claw-Eval Avg | coding_agent | 68.7 | 53.2 | 65.4 |
| BrowseComp | general_agent | 62 | 8.3 | 61 |
| MCPMark | general_agent | 37 | 13.4 | 27 |
| SWE-bench Multilingual | coding_agent | 67.2 | 43.3 | 60.3 |
| IFBench | instruction_following | 66.6 | 64.5 | 70.2 |
| SWE-bench Verified | coding_agent | 73.4 | 53.1 | 70 |
| IMOAnswerBench | stem_reasoning | 78.9 | 69.8 | 76.8 |
| Humanity's Last Exam | stem_reasoning | 21.4 | 14.3 | 22.4 |
| Terminal-Bench 2.1 (Terminus-2) | coding_agent | 52.5 | 21.3 | 41.4 |
| HMMT Feb 26 | stem_reasoning | 83.6 | 70.8 | 78.7 |
| MCP-Atlas | general_agent | 62.8 | 47.4 | 62.4 |
| SWE Atlas - RF | coding_agent | 11.4 | 4.3 | 10.2 |
| Terminal-Bench 2.1 (Claude Code) | coding_agent | 49.2 | 18.9 | 38.9 |
Partial coverage benchmarks
Scored by at least one but not all selected models (108) — missing scores show as —
| Benchmark | Category | Qwen3.6-35B-A3B | Qwen3.5-9B | Qwen3.5-35B-A3B |
|---|---|---|---|---|
| HMMT Feb 25 2/3 models | stem_reasoning | 90.7 | — | 89 |
| DynaMath 2/3 models | vision_language | 82.8 | — | 85 |
| GPQA Diamond 2/3 models | stem_reasoning | 86 | — | 84.2 |
| MathVista (mini) 2/3 models | vision_language | 86.4 | — | 86.2 |
| MMBench EN-DEV v1.1 2/3 models | vision_language | 92.8 | — | 91.5 |
| Nuscene 1/3 models | spatial_intelligence | — | — | 14.6 |
| LiveCodeBench v6 2/3 models | stem_reasoning | 80.4 | — | 74.6 |
| RealWorldQA 2/3 models | vision_language | 85.3 | — | 84.1 |
| EmbSpatialBench 2/3 models | spatial_intelligence | 84.3 | — | 83.1 |
| Hypersim 1/3 models | spatial_intelligence | — | — | 13.1 |
| MathVision 1/3 models | vision_language | — | — | 83.9 |
| BrowseComp-zh 1/3 models | general_agent | — | — | 69.5 |
| MAXIFE 1/3 models | multilingual | — | — | 86.6 |
| AndroidWorld 1/3 models | general_agent | — | — | 71.1 |
| AgentWorldBench Terminal 1/3 models | general_agent | — | — | 46.13 |
| Toolathlon Verified 1/3 models | general_agent | 41.7 | — | — |
| ScreenSpot Pro 1/3 models | general_agent | — | — | 68.6 |
| OJBench 1/3 models | stem_reasoning | — | — | 36 |
| AgentWorldBench Overall 1/3 models | general_agent | — | — | 47.73 |
| AutomationBench (Public) 1/3 models | general_agent | 24.9 | — | — |
| MMStar 2/3 models | vision_language | 80.7 | — | 81.9 |
| GDPVal 1/3 models | general_agent | 67.4 | — | — |
| OCRBench 1/3 models | document_understanding | — | — | 91 |
| Seal-0 1/3 models | general_agent | — | — | 41.4 |
| PMC-VQA 1/3 models | vision_language | — | — | 62 |
| MMLongBench-Doc 1/3 models | document_understanding | — | — | 59.5 |
| WildClawBench 1/3 models | coding_agent | 68.7 | — | — |
| AgentWorldBench Web 1/3 models | general_agent | — | — | 47.1 |
| AI2D_TEST 2/3 models | document_understanding | 92.7 | — | 92.6 |
| LingoQA 1/3 models | spatial_intelligence | — | — | 79.2 |
| ParseBench Text Formatting 1/3 models | document_understanding | 58.3 | — | — |
| NOVA-63 1/3 models | multilingual | — | — | 57.1 |
| MMMU-Pro 2/3 models | vision_language | 75.3 | — | 75.1 |
| INCLUDE 1/3 models | multilingual | — | — | 79.7 |
| QwenClawBench 2/3 models | coding_agent | 52.6 | — | 47.7 |
| BabyVision 1/3 models | vision_language | — | — | 38.4 |
| SLAKE 1/3 models | vision_language | — | — | 78.7 |
| MMMLU 1/3 models | multilingual | — | — | 85.2 |
| VideoMMMU 2/3 models | video_understanding | 83.7 | — | 80.4 |
| HMMT Nov 25 2/3 models | stem_reasoning | 89.1 | — | 89.2 |
| Multi-Challenge 2/3 models | instruction_following | 58.2 | — | 60 |
| PolyMATH 1/3 models | multilingual | — | — | 64.4 |
| MMLU-ProX 1/3 models | multilingual | — | — | 81 |
| MMLU-Pro 2/3 models | knowledge | 85.2 | — | 85.3 |
| CodeForces 1/3 models | stem_reasoning | — | — | 2028 |
| SuperGPQA 2/3 models | knowledge | 64.7 | — | 63.4 |
| OSWorld-Verified 1/3 models | general_agent | — | — | 54.5 |
| TIR-Bench 1/3 models | vision_language | — | — | 55.5 |
| Global PIQA 1/3 models | multilingual | — | — | 86.6 |
| AgentWorldBench Search 1/3 models | general_agent | — | — | 25.98 |
| IFEval 2/3 models | instruction_following | — | 91.5 | 91.9 |
| HLE (with tools) 2/3 models | stem_reasoning | 28.9 | — | 47.4 |
| LVBench 2/3 models | video_understanding | 71.4 | — | 71.4 |
| CharXiv (RQ) 2/3 models | document_understanding | 78 | — | 77.5 |
| GDP.pdf 1/3 models | long_context | 6.6 | — | — |
| MedXpertQA-MM 1/3 models | vision_language | — | — | 61.4 |
| MMLU-Redux 2/3 models | knowledge | 93.3 | — | 93.3 |
| ParseBench Text Content 1/3 models | document_understanding | 90.7 | — | — |
| AA-Omniscience Index (public set) 1/3 models | knowledge | -7.2 | — | — |
| MMVU 1/3 models | video_understanding | — | — | 72.3 |
| FullStackBench zh 1/3 models | coding_agent | — | — | 55 |
| MLVU 2/3 models | video_understanding | 86.2 | — | 85.6 |
| FullStackBench en 1/3 models | coding_agent | — | — | 58.1 |
| CountBench 2/3 models | spatial_intelligence | 96.1 | — | 97.8 |
| ERQA 2/3 models | spatial_intelligence | 61.8 | — | 64.8 |
| VideoMME (w/o sub.) 2/3 models | video_understanding | 82.5 | — | 82.5 |
| Terminal-Bench 2.0 2/3 models | coding_agent | 51.5 | — | 40.5 |
| DeepPlanning 2/3 models | general_agent | 25.9 | — | 22.8 |
| QwenWebBench 2/3 models | coding_agent | 1397 | — | 978 |
| TAU2-Bench 2/3 models | general_agent | — | 79.1 | 81.2 |
| ODInW13 2/3 models | spatial_intelligence | 50.8 | — | 42.6 |
| VlmsAreBlind 2/3 models | vision_language | 96.6 | — | 97 |
| DeepSWE 1/3 models | coding_agent | 0 | — | — |
| OmniDocBench 1.5 2/3 models | document_understanding | 89.9 | — | 89.3 |
| HallusionBench 2/3 models | vision_language | 69.8 | — | 67.9 |
| OCRBench 2/3 models | document_understanding | 90 | — | 89.4 |
| Terminal-Bench Lite 1/3 models | coding_agent | 63.3 | — | — |
| AgentWorldBench OS 1/3 models | general_agent | — | — | 56.27 |
| AgentWorldBench MCP 1/3 models | general_agent | — | — | 57.87 |
| GPQA 1/3 models | reasoning | — | 77.2 | — |
| AgentWorldBench SWE 1/3 models | general_agent | — | — | 47.58 |
| CC-OCR 2/3 models | document_understanding | 81.9 | — | 80.7 |
| RefCOCO (avg) 2/3 models | spatial_intelligence | 92 | — | 89.2 |
| Artificial Analysis Intelligence Index 1/3 models | composite | 32 | — | — |
| Tool Decathlon 2/3 models | general_agent | 26.9 | — | 28.7 |
| Frontier-Bench v0.1 1/3 models | coding_agent | 1.4 | — | — |
| ParseBench Mean 1/3 models | document_understanding | 44.1 | — | — |
| ZEROBench_sub 2/3 models | vision_language | 34.4 | — | 34.1 |
| C-Eval 2/3 models | knowledge | 90 | — | 90.2 |
| WideSearch 2/3 models | general_agent | 60.1 | — | 59.1 |
| SimpleVQA 2/3 models | vision_language | 58.9 | — | 58.3 |
| VideoMME (w sub.) 2/3 models | video_understanding | 86.6 | — | 86.6 |
| Workspace Bench 1/3 models | — | — | 25.5 | — |
| SkillsBench Avg5 2/3 models | coding_agent | 28.7 | — | 4.4 |
| Claw-Eval Pass^3 2/3 models | coding_agent | 50 | — | 51 |
| WMT24++ (en→xx) 1/3 models | multilingual | — | — | 76.3 |
| LongBench v2 1/3 models | long_context | — | — | 59 |
| V-Star 2/3 models | vision_language | 90.1 | — | 92.7 |
| AgentWorldBench Android 1/3 models | general_agent | — | — | 53.18 |
| Chatbot Korean (internal) 1/3 models | multilingual | 74.5 | — | — |
| Gaokao 2026 1/3 models | — | — | 135.5 | — |
| SUNRGBD 1/3 models | spatial_intelligence | — | — | 33.4 |
| SciCode 2/3 models | — | 35.6 | 32.7 | — |
| RefSpatialBench 2/3 models | spatial_intelligence | 64.3 | — | 63.5 |
| BrowseComp Plus 1/3 models | general_agent | 34.8 | — | — |
| MMMU 2/3 models | vision_language | 81.7 | — | 81.4 |
| MVBench 2/3 models | video_understanding | 74.6 | — | 74.8 |
| ZEROBench 1/3 models | vision_language | — | — | 8 |
Comparison FAQ
Which model scores highest in this comparison?
Qwen3.6-35B-A3B wins this comparison with an overall score of 86 vs 67 for Qwen3.5-35B-A3B, leading 18 of 22 shared benchmarks.
Which model is strongest at Agentic Coding?
Qwen3.6-35B-A3B leads the Agentic Coding category (strength score 99.7/100 across 15 comparable benchmarks).
How reliable is this comparison verdict?
Based on 22 of 130 benchmarks scored by all 3 selected models.