Alibaba Cloud / Qwen Team model product
5-27B is a multimodal dense foundation model with 27 billion parameters.
Updated Aug 12, 2026. Default version: Qwen3.5-27B
Technical details for the model's default version.
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for Qwen3.5-27B.
| CountBench | 0.978 points | 01 | 6 | 100.0% | C | |
| Hallusion Bench | 70.0% | 01 | 16 | 100.0% | C | |
| IFEval | 95.0% | 01 | 65 | 100.0% | C | |
| DynaMath | 87.7% | 02 | 7 | 83.3% | C | |
| EmbSpatialBench | 0.845 points | 02 | 8 | 85.7% | C | |
| FullStackBench en | 60.1% | 02 | 3 | 50.0% | C | |
| FullStackBench zh | 57.4% | 02 | 3 | 50.0% | C | |
| Hypersim | 0.13 points | 02 | 4 | 66.7% | C | |
| LingoQA | 82.0% | 02 | 4 | 66.7% | C | |
| MathVista-Mini | 87.8% | 02 | 23 | 95.5% | C | |
| MMLongBench-Doc | 0.602 points | 02 | 5 | 75.0% | C | |
| Nuscene | 15.2% | 02 | 3 | 50.0% | C | |
| PMC-VQA | 62.4% | 02 | 3 | 50.0% | C | |
| SlakeVQA | 80.0% | 02 | 4 | 66.7% | C | |
| SUNRGBD | 0.354 points | 02 | 4 | 66.7% | C | |
| TIR-Bench | 59.8% | 02 | 4 | 66.7% | C | |
| VideoMME w/o sub. | 82.8% | 02 | 10 | 88.9% | C | |
| ZEROBench-Sub | 0.362 points | 02 | 5 | 75.0% | C | |
| AndroidWorld_SR | 64.2% | 03 | 8 | 71.4% | C | |
| MedXpertQA | 62.4% | 03 | 12 | 81.8% | C | |
| MMBench-V1.1 | 92.6% | 03 | 18 | 88.2% | C | |
| MMVU | 73.3% | 03 | 4 | 33.3% | C | |
| OJBench | 40.1% | 03 | 9 | 75.0% | C | |
| Seal-0 | 47.2% | 03 | 6 | 60.0% | C | |
| VLMsAreBlind | 96.9% | 03 | 4 | 33.3% | C | |
| MVBench | 74.6% | 04 | 17 | 81.3% | C | |
| RefSpatialBench | 0.677 points | 04 | 6 | 40.0% | C | |
| V* | 93.7% | 04 | 7 | 50.0% | C | |
| VideoMME w sub. | 87.0% | 04 | 10 | 66.7% | C | |
| BFCL-V4 | 68.5% | 05 | 13 | 66.7% | C | |
| MAXIFE | 88.0% | 05 | 11 | 60.0% | C | |
| MMStar | 81.0% | 05 | 22 | 81.0% | C | |
| NOVA-63 | 58.1% | 05 | 11 | 60.0% | C | |
| PolyMATH | 71.2% | 05 | 23 | 81.8% | C | |
| VITA-Bench | 41.9% | 05 | 10 | 55.6% | C | |
| MLVU | 85.9% | 06 | 10 | 44.4% | C | |
| RefCOCO-avg | 0.909 points | 06 | 7 | 16.7% | C | |
| WideSearch | 61.1% | 06 | 9 | 37.5% | C | |
| ZEROBench | 0.1 points | 06 | 9 | 37.5% | C | |
| AI2D | 92.9% | 07 | 32 | 80.7% | C | |
| BabyVision | 44.6% | 07 | 9 | 25.0% | C | |
| CC-OCR | 81.0% | 07 | 18 | 64.7% | C | |
| C-Eval | 90.5% | 07 | 18 | 64.7% | C | |
| CodeForces | 80.7% | 07 | 16 | 60.0% | C | |
| DeepPlanning | 22.6% | 07 | 9 | 25.0% | C | |
| HMMT25 | 89.8% | 07 | 25 | 75.0% | C | |
| Include | 81.6% | 07 | 31 | 80.0% | C | |
| LVBench | 73.6% | 07 | 24 | 73.9% | C | |
| MMLU-ProX | 82.2% | 07 | 32 | 80.7% | C | |
| AA-LCR | 66.1% | 08 | 16 | 53.3% | C | |
| Global PIQA | 87.5% | 08 | 13 | 41.7% | C | |
| IFBench | 76.5% | 08 | 29 | 75.0% | C | |
| MathVision | 86.0% | 08 | 32 | 77.4% | C | |
| OCRBench | 89.4% | 08 | 22 | 66.7% | C | |
| OmniDocBench 1.5 | 88.9% | 08 | 17 | 56.3% | C | |
| BrowseComp-zh | 62.1% | 09 | 13 | 33.3% | C | |
| MMMU | 82.3% | 09 | 63 | 87.1% | C | |
| Multi-Challenge | 60.8% | 09 | 29 | 71.4% | C | |
| ScreenSpot Pro | 70.3% | 09 | 25 | 66.7% | C | |
| SuperGPQA | 65.6% | 09 | 34 | 75.8% | C | |
| LongBench v2 | 60.6% | 10 | 17 | 43.8% | C | |
| RealWorldQA | 83.7% | 10 | 26 | 64.0% | C | |
| WMT24++ | 77.6% | 10 | 23 | 59.1% | C | |
| ERQA | 60.5% | 12 | 23 | 50.0% | C | |
| MMLU-Pro | 86.1% | 13 | 129 | 90.6% | C | |
| MMLU-Redux | 93.2% | 13 | 48 | 74.5% | C | |
| SimpleVQA | 0.56 points | 13 | 13 | 0.0% | C | |
| HMMT 2025 | 92.0% | 14 | 33 | 59.4% | C | |
| ODinW | 41.1% | 14 | 16 | 13.3% | C | |
| VideoMMMU | 82.3% | 14 | 26 | 48.0% | C | |
| t2-bench | 79.0% | 15 | 23 | 36.4% | C | |
| LiveCodeBench v6 | 80.7% | 18 | 53 | 67.3% | C | |
| CharXiv-R | 79.5% | 21 | 48 | 57.5% | C | |
| OSWorld-Verified | 56.2% | 21 | 23 | 9.1% | C | |
| Humanity's Last Exam | 48.5% | 22 | 93 | 77.2% | C | |
| MMMLU | 85.9% | 26 | 49 | 47.9% | C | |
| BrowseComp | 61.0% | 34 | 58 | 42.1% | C | |
| MMMU-Pro | 75.0% | 34 | 66 | 49.2% | C | |
| Terminal-Bench 2.0 | 41.6% | 43 | 49 | 12.5% | C | |
| GPQA | 85.5% | 52 | 234 | 78.1% | C | |
| SWE-Bench Verified | 72.4% | 52 | 105 | 51.0% | C |
Preference and agent-evaluation results for the default version.
| vision | overall | 55 | 1240.9 | 21,450 | N/A | |
| vision style control | overall | 59 | 1218.7 | 21,450 | N/A | |
| webdev | overall | 78 | 1357.2 | 7,428 | N/A | |
| text factuality | overall | 109 | 1418.9 | 27,160 | N/A | |
| text | overall | 122 | 1407.9 | 27,339 | N/A | |
| text style control | overall | 130 | 1408.0 | 27,339 | N/A |
Provider-specific output speed and catalog latency for Qwen3.5-27B. Runtime does not affect the capability score.
| Novita | 7.909 tok/s | 6.349 s | 262.1K | 65.5K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Official vendor API pricing appears first, followed by individual provider offers.
| EmpirioLabs AI | qwen3-5-27b | global | $0.086 | $0.688 | 256K | |
| OrcaRouter | qwen/qwen3.5-27b | global | $0.086 | $0.688 | 262.1K | |
| Merge Gateway | qwen/qwen3.5-27b | global | $0.086 | $0.688 | 256K | |
| Kilo Gateway | qwen/qwen3.5-27b | global | $0.195 | $1.56 | 262.1K | |
| OpenRouter | qwen/qwen3.5-27b | global | $0.195 | $1.56 | 262.1K | |
| SiliconFlow | Qwen/Qwen3.5-27B | global | $0.25 | $2 | 262.1K | |
| Deep Infra | Qwen/Qwen3.5-27B | global | $0.26 | $2.6 | 262.1K | |
| NanoGPT | qwen3.5-27b | global | $0.27 | $2.16 | 260.1K | |
| Ofox | bailian/qwen3.5-27b | global | $0.29 | $2.05 | 262.1K | |
| Mixlayer | qwen/qwen3.5-27b | global | $0.30 | $2.4 | 262.1K | |
| Alibaba | qwen3.5-27b | global | $0.30 | $2.4 | 262.1K | |
| NovitaAI | qwen/qwen3.5-27b | global | $0.30 | $2.4 | 262.1K | |
| Hugging Face | Qwen/Qwen3.5-27B | global | $0.30 | $2.4 | 262.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Available versions of this model. The score column identifies the version used in the overall ranking.
| Qwen3.5-27B | 57.2 | 27B | 262.1K | 65.5K | Yes | Apache 2.0 |
Key information about Qwen3.5 27B and its available data.
5-27B is a multimodal dense foundation model with 27 billion parameters. It combines strong reasoning, coding, multilingual, long-context, and visual understanding performance in a production-friendly open-weight package with a native 262K context window.
Data as of 2026-08-11.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Common questions about Qwen3.5 27B.
Qwen3.5 27B's default version was released on Feb 24, 2026.
Qwen3.5 27B's official API price is $0.30 per million input tokens and $2.4 per million output tokens via Alibaba. The lowest tracked third-party offer starts at $0.086 input and $0.688 output via EmpirioLabs AI.
Qwen3.5 27B was created by Alibaba Cloud / Qwen Team.
The default version has a 262.1K token context window.
Yes. The default version is marked as open weight under Apache 2.0.
13 provider offerings are linked to the default version.
Nearby ranked alternatives include Muse Glimmer 30B, MiniMax M2.7, DeepSeek-V3.2.