Qwen3.6-27B vs Qwen3.6-35B-A3B
New ComparisonScores are fetched live from the current database on every visit — nothing is cached.
Winner
Qwen3.6-27B
Qwen3.6-27B wins this comparison with an overall score of 88 vs 66 for Qwen3.6-35B-A3B, leading 42 of 50 shared benchmarks.
Qwen3.6-27B
88.3
Qwen3.6-35B-A3B
65.8
Based on 50 of 113 benchmarks scored by all 2 selected models.
Benchmark Scores Comparison
Qwen3.6-35B-A3B
Qwen3.6-27B
Qwen3.6-35B-A3B
35.0B params
Score: 100.0
Qwen3.6-27B
27.0B params
Score: 100.0
Coverage matrix
2 models × 113 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 113 benchmarks each model has scores for.
| Benchmark | Qwen3.6-35B-A3B Qwen Team (Alibaba Cloud) | Qwen3.6-27B Qwen Team (Alibaba Cloud) |
|---|---|---|
| HMMT Feb 25 | 90.7 | 93.8 |
| DynaMath | 82.8 | 85.6 |
| GPQA Diamond | 86 | 87.8 |
| IFBench (prompt loose) | — | 69.1 |
| MathVista (mini) | 86.4 | 87.4 |
| MMBench EN-DEV v1.1 | 92.8 | 92.3 |
| LiveCodeBench v6 | 80.4 | 83.9 |
| SWE-bench Pro | 49.5 | 53.5 |
| BFCL-V4 | 64.7 | — |
| RealWorldQA | 85.3 | 84.1 |
| EmbSpatialBench | 84.3 | 84.6 |
| AA-LCR | 69.3 | 73.3 |
| MathVision | — | 85.1 |
| AIME 26 | 92.7 | 94.1 |
| AndroidWorld | — | 70.3 |
| Agents' Last Exam | — | 10.6 |
| Toolathlon Verified | 41.7 | — |
| RecreationBench | — | 29.8 |
| VITA-Bench | 35.6 | — |
| TAU3-Bench | 67.2 | — |
| ScreenSpot Pro | — | 76.1 |
| WMDP (Bio) | — | 84.8 |
| AutomationBench (Public) | 24.9 | — |
| MMStar | 80.7 | 81.4 |
| GDPVal | 67.4 | — |
| DeepSearch QA | — | 71.1 |
| SWE Atlas - TW | 13.3 | — |
| WildClawBench | 68.7 | 43.2 |
| SWE Atlas - QnA | 15.5 | — |
| AI2D_TEST | 92.7 | — |
| ParseBench Text Formatting | 58.3 | — |
| NL2Repo | 29.4 | 36.2 |
| HPCT | — | 48.7 |
| MMMU-Pro | 75.3 | 75.8 |
| QwenClawBench | 52.6 | 53.4 |
| BabyVision | — | 28.9 |
| Claw-Eval Avg | 68.7 | 72.4 |
| VideoMMMU | 83.7 | 84.4 |
| HMMT Nov 25 | 89.1 | 90.7 |
| Multi-Challenge | 58.2 | — |
| SWE-MM | — | 25.7 |
| MMLU-Pro | 85.2 | 86.2 |
| SuperGPQA | 64.7 | 66 |
| OSWorld-Verified | — | 63.9 |
| HLE (with tools) | 28.9 | — |
| LVBench | 71.4 | — |
| CharXiv (RQ) | 78 | 78.4 |
| GDP.pdf | 6.6 | — |
| BrowseComp | 62 | — |
| MMLU-Redux | 93.3 | 93.5 |
| SkillsBench | — | 46.6 |
| WebArena-Verified | — | 48.8 |
| ParseBench Text Content | 90.7 | — |
| CoWorkBench | — | 61 |
| Lab Bench (ProtocolQA) | — | 69.1 |
| MCPMark | 37 | — |
| AA-Omniscience Index (public set) | -7.2 | — |
| SWE-bench Multilingual | 67.2 | 71.3 |
| MLVU | 86.2 | 86.6 |
| Beam128K | — | 63 |
| IFBench | 66.6 | 70.8 |
| SWE-bench Verified | 73.4 | 77.2 |
| IMOAnswerBench | 78.9 | 80.8 |
| Humanity's Last Exam | 21.4 | 24 |
| CountBench | 96.1 | 97.8 |
| WMDP (Chem) | — | 74.8 |
| ERQA | 61.8 | 62.5 |
| VideoMME (w/o sub.) | 82.5 | — |
| Terminal-Bench 2.0 | 51.5 | 59.3 |
| DeepPlanning | 25.9 | — |
| QwenSWEBench | — | 49.3 |
| QwenWebBench | 1397 | 1487 |
| ODInW13 | 50.8 | — |
| VlmsAreBlind | 96.6 | 97 |
| Terminal-Bench 2.1 (Terminus-2) | 52.5 | 63.4 |
| DeepSWE | 0 | — |
| OmniDocBench 1.5 | 89.9 | 89.4 |
| HallusionBench | 69.8 | — |
| OCRBench | 90 | 89.4 |
| HMMT Feb 26 | 83.6 | 84.3 |
| Terminal-Bench Lite | 63.3 | — |
| CC-OCR | 81.9 | 81.2 |
| RefCOCO (avg) | 92 | 92.5 |
| JobBench | — | 21.8 |
| Artificial Analysis Intelligence Index | 32 | — |
| SciCode (subtask) | — | 39.8 |
| Tool Decathlon | 26.9 | — |
| Frontier-Bench v0.1 | 1.4 | — |
| ParseBench Mean | 44.1 | — |
| ZEROBench_sub | 34.4 | — |
| C-Eval | 90 | 91.4 |
| WideSearch | 60.1 | — |
| SimpleVQA | 58.9 | 56.1 |
| VideoMME (w sub.) | 86.6 | 87.7 |
| SkillsBench Avg5 | 28.7 | 48.2 |
| Vision2Web | — | 45 |
| TauBench V3 Banking | — | 16.7 |
| Claw-Eval Pass^3 | 50 | 60.6 |
| MCP-Atlas | 62.8 | 62.5 |
| SWE Atlas - RF | 11.4 | — |
| V-Star | 90.1 | 94.7 |
| DeepSWE 1.1 | — | 13.3 |
| Terminal-Bench 2.1 (Claude Code) | 49.2 | — |
| Chatbot Korean (internal) | 74.5 | — |
| MBCT | — | 45.9 |
| VCT | — | 33.7 |
| Gaia2 | — | 40 |
| SciCode | 35.6 | — |
| RefSpatialBench | 64.3 | 70 |
| BrowseComp Plus | 34.8 | — |
| MMMU | 81.7 | 82.9 |
| GDPVal-AA v2 | — | 1141 |
| MVBench | 74.6 | 75.5 |
| Covered | 85/113 | 78/113 |
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 |
|---|---|---|---|---|---|
| Qwen3.6-27B Winner | 88.3 | 42 of 50 | 78 of 113 (69.0%) | Video Understanding | Agentic Coding |
| Qwen3.6-35B-A3B | 65.8 | 8 of 50 | 85 of 113 (75.0%) | General Agents | 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 | Qwen3.6-27B | Qwen3.6-35B-A3B |
|---|---|---|
| General Agents 26 benchmarks | 99.4 | 100 |
| Agentic Coding 22 benchmarks | 96.7 | 83.5 |
| Vision & Language 15 benchmarks | 98.2 | 93.4 |
| STEM & Reasoning 10 benchmarks | 100 | 94.2 |
| Document Understanding 8 benchmarks | 97.5 | 99.2 |
| Spatial Intelligence 6 benchmarks | 100 | 92.4 |
| Video Understanding 6 benchmarks | 100 | 96.1 |
| Knowledge 5 benchmarks | 100 | 94.2 |
| Instruction Following 3 benchmarks | 100 | 91.3 |
| Long Context 3 benchmarks | 100 | 94.5 |
Fully comparable benchmarks
Scored by all 2 selected models (50)
| Benchmark | Category | Qwen3.6-35B-A3B | Qwen3.6-27B |
|---|---|---|---|
| HMMT Feb 25 | stem_reasoning | 90.7 | 93.8 |
| DynaMath | vision_language | 82.8 | 85.6 |
| GPQA Diamond | stem_reasoning | 86 | 87.8 |
| MathVista (mini) | vision_language | 86.4 | 87.4 |
| MMBench EN-DEV v1.1 | vision_language | 92.8 | 92.3 |
| LiveCodeBench v6 | stem_reasoning | 80.4 | 83.9 |
| SWE-bench Pro | coding_agent | 49.5 | 53.5 |
| RealWorldQA | vision_language | 85.3 | 84.1 |
| EmbSpatialBench | spatial_intelligence | 84.3 | 84.6 |
| AA-LCR | long_context | 69.3 | 73.3 |
| AIME 26 | stem_reasoning | 92.7 | 94.1 |
| MMStar | vision_language | 80.7 | 81.4 |
| WildClawBench | coding_agent | 68.7 | 43.2 |
| NL2Repo | coding_agent | 29.4 | 36.2 |
| MMMU-Pro | vision_language | 75.3 | 75.8 |
| QwenClawBench | coding_agent | 52.6 | 53.4 |
| Claw-Eval Avg | coding_agent | 68.7 | 72.4 |
| VideoMMMU | video_understanding | 83.7 | 84.4 |
| HMMT Nov 25 | stem_reasoning | 89.1 | 90.7 |
| MMLU-Pro | knowledge | 85.2 | 86.2 |
| SuperGPQA | knowledge | 64.7 | 66 |
| CharXiv (RQ) | document_understanding | 78 | 78.4 |
| MMLU-Redux | knowledge | 93.3 | 93.5 |
| SWE-bench Multilingual | coding_agent | 67.2 | 71.3 |
| MLVU | video_understanding | 86.2 | 86.6 |
| IFBench | instruction_following | 66.6 | 70.8 |
| SWE-bench Verified | coding_agent | 73.4 | 77.2 |
| IMOAnswerBench | stem_reasoning | 78.9 | 80.8 |
| Humanity's Last Exam | stem_reasoning | 21.4 | 24 |
| CountBench | spatial_intelligence | 96.1 | 97.8 |
| ERQA | spatial_intelligence | 61.8 | 62.5 |
| Terminal-Bench 2.0 | coding_agent | 51.5 | 59.3 |
| QwenWebBench | coding_agent | 1397 | 1487 |
| VlmsAreBlind | vision_language | 96.6 | 97 |
| Terminal-Bench 2.1 (Terminus-2) | coding_agent | 52.5 | 63.4 |
| OmniDocBench 1.5 | document_understanding | 89.9 | 89.4 |
| OCRBench | document_understanding | 90 | 89.4 |
| HMMT Feb 26 | stem_reasoning | 83.6 | 84.3 |
| CC-OCR | document_understanding | 81.9 | 81.2 |
| RefCOCO (avg) | spatial_intelligence | 92 | 92.5 |
| C-Eval | knowledge | 90 | 91.4 |
| SimpleVQA | vision_language | 58.9 | 56.1 |
| VideoMME (w sub.) | video_understanding | 86.6 | 87.7 |
| SkillsBench Avg5 | coding_agent | 28.7 | 48.2 |
| Claw-Eval Pass^3 | coding_agent | 50 | 60.6 |
| MCP-Atlas | general_agent | 62.8 | 62.5 |
| V-Star | vision_language | 90.1 | 94.7 |
| RefSpatialBench | spatial_intelligence | 64.3 | 70 |
| MMMU | vision_language | 81.7 | 82.9 |
| MVBench | video_understanding | 74.6 | 75.5 |
Partial coverage benchmarks
Scored by at least one but not all selected models (63) — missing scores show as —
| Benchmark | Category | Qwen3.6-35B-A3B | Qwen3.6-27B |
|---|---|---|---|
| IFBench (prompt loose) 1/2 models | instruction_following | — | 69.1 |
| BFCL-V4 1/2 models | general_agent | 64.7 | — |
| MathVision 1/2 models | vision_language | — | 85.1 |
| AndroidWorld 1/2 models | general_agent | — | 70.3 |
| Agents' Last Exam 1/2 models | general_agent | — | 10.6 |
| Toolathlon Verified 1/2 models | general_agent | 41.7 | — |
| RecreationBench 1/2 models | general_agent | — | 29.8 |
| VITA-Bench 1/2 models | general_agent | 35.6 | — |
| TAU3-Bench 1/2 models | general_agent | 67.2 | — |
| ScreenSpot Pro 1/2 models | general_agent | — | 76.1 |
| WMDP (Bio) 1/2 models | safety | — | 84.8 |
| AutomationBench (Public) 1/2 models | general_agent | 24.9 | — |
| GDPVal 1/2 models | general_agent | 67.4 | — |
| DeepSearch QA 1/2 models | general_agent | — | 71.1 |
| SWE Atlas - TW 1/2 models | coding_agent | 13.3 | — |
| SWE Atlas - QnA 1/2 models | coding_agent | 15.5 | — |
| AI2D_TEST 1/2 models | document_understanding | 92.7 | — |
| ParseBench Text Formatting 1/2 models | document_understanding | 58.3 | — |
| HPCT 1/2 models | safety | — | 48.7 |
| BabyVision 1/2 models | vision_language | — | 28.9 |
| Multi-Challenge 1/2 models | instruction_following | 58.2 | — |
| SWE-MM 1/2 models | coding_agent | — | 25.7 |
| OSWorld-Verified 1/2 models | general_agent | — | 63.9 |
| HLE (with tools) 1/2 models | stem_reasoning | 28.9 | — |
| LVBench 1/2 models | video_understanding | 71.4 | — |
| GDP.pdf 1/2 models | long_context | 6.6 | — |
| BrowseComp 1/2 models | general_agent | 62 | — |
| SkillsBench 1/2 models | general_agent | — | 46.6 |
| WebArena-Verified 1/2 models | general_agent | — | 48.8 |
| ParseBench Text Content 1/2 models | document_understanding | 90.7 | — |
| CoWorkBench 1/2 models | general_agent | — | 61 |
| Lab Bench (ProtocolQA) 1/2 models | safety | — | 69.1 |
| MCPMark 1/2 models | general_agent | 37 | — |
| AA-Omniscience Index (public set) 1/2 models | knowledge | -7.2 | — |
| Beam128K 1/2 models | long_context | — | 63 |
| WMDP (Chem) 1/2 models | safety | — | 74.8 |
| VideoMME (w/o sub.) 1/2 models | video_understanding | 82.5 | — |
| DeepPlanning 1/2 models | general_agent | 25.9 | — |
| QwenSWEBench 1/2 models | coding_agent | — | 49.3 |
| ODInW13 1/2 models | spatial_intelligence | 50.8 | — |
| DeepSWE 1/2 models | coding_agent | 0 | — |
| HallusionBench 1/2 models | vision_language | 69.8 | — |
| Terminal-Bench Lite 1/2 models | coding_agent | 63.3 | — |
| JobBench 1/2 models | general_agent | — | 21.8 |
| Artificial Analysis Intelligence Index 1/2 models | composite | 32 | — |
| SciCode (subtask) 1/2 models | stem_reasoning | — | 39.8 |
| Tool Decathlon 1/2 models | general_agent | 26.9 | — |
| Frontier-Bench v0.1 1/2 models | coding_agent | 1.4 | — |
| ParseBench Mean 1/2 models | document_understanding | 44.1 | — |
| ZEROBench_sub 1/2 models | vision_language | 34.4 | — |
| WideSearch 1/2 models | general_agent | 60.1 | — |
| Vision2Web 1/2 models | vision_language | — | 45 |
| TauBench V3 Banking 1/2 models | general_agent | — | 16.7 |
| SWE Atlas - RF 1/2 models | coding_agent | 11.4 | — |
| DeepSWE 1.1 1/2 models | coding_agent | — | 13.3 |
| Terminal-Bench 2.1 (Claude Code) 1/2 models | coding_agent | 49.2 | — |
| Chatbot Korean (internal) 1/2 models | multilingual | 74.5 | — |
| MBCT 1/2 models | safety | — | 45.9 |
| VCT 1/2 models | safety | — | 33.7 |
| Gaia2 1/2 models | general_agent | — | 40 |
| SciCode 1/2 models | — | 35.6 | — |
| BrowseComp Plus 1/2 models | general_agent | 34.8 | — |
| GDPVal-AA v2 1/2 models | general_agent | — | 1141 |
Comparison FAQ
Which model is better, Qwen3.6-27B or Qwen3.6-35B-A3B?
Qwen3.6-27B wins this comparison with an overall score of 88 vs 66 for Qwen3.6-35B-A3B, leading 42 of 50 shared benchmarks.
Which model is better at Agentic Coding, Qwen3.6-27B or Qwen3.6-35B-A3B?
Qwen3.6-27B leads the Agentic Coding category (strength score 96.7/100 across 12 comparable benchmarks).
Where does Qwen3.6-35B-A3B outperform Qwen3.6-27B?
Qwen3.6-35B-A3B leads the Document Understanding category (strength score 99.2/100 across 4 comparable benchmarks).
How reliable is this comparison verdict?
Based on 50 of 113 benchmarks scored by all 2 selected models.