Alibaba Cloud / Qwen Team model product
Qwen3.6 27B
6 series, supporting both vision-language thinking and non-thinking modes in a single unified checkpoint.
Updated Aug 12, 2026. Default version: Qwen3.6-27B
LLMBoard score60.5Qwen3.6-27B
Coverage80%43 benchmark families
Context window262.1KTokens
Official input price$0.60Alibaba API
Qwen3.6 27B Specifications
Technical details for the model's default version.
- Version
- Qwen3.6-27B
- Released
- Apr 21, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 27.8B
- Context window
- 262.1K
- Max output
- 65.5K
- Inputs
- audio, image, text, video
- Outputs
- text
- Open weights
- Yes
- License
- Apache 2.0
Qwen3.6 27B Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Qwen3.6 27B Benchmark Results
Benchmark scores for Qwen3.6-27B.
| EmbSpatialBench | 0.846 points | 01 | 8 | 100.0% | C | |
| RefSpatialBench | 0.7 points | 01 | 6 | 100.0% | C | |
| MVBench | 75.5% | 02 | 17 | 93.8% | C | |
| QwenWebBench | 1,487 points | 02 | 2 | 0.0% | C | |
| RefCOCO-avg | 0.925 points | 02 | 7 | 83.3% | C | |
| VideoMME w sub. | 87.7% | 02 | 10 | 88.9% | C | |
| VLMsAreBlind | 97.0% | 02 | 4 | 66.7% | C | |
| CountBench | 0.978 points | 03 | 6 | 60.0% | C | |
| V* | 94.7% | 03 | 7 | 66.7% | C | |
| ZClawBench | 53.4% | 03 | 4 | 33.3% | C | |
| AndroidWorld | 70.3% | 04 | 4 | 0.0% | C | |
| DynaMath | 85.6% | 04 | 7 | 50.0% | C | |
| HMMT25 | 90.7% | 04 | 25 | 87.5% | C | |
| MathVista-Mini | 87.4% | 04 | 23 | 86.4% | C | |
| MLVU | 86.6% | 04 | 10 | 66.7% | C | |
| MMBench-V1.1 | 92.3% | 04 | 18 | 82.3% | C | |
| MMStar | 81.4% | 04 | 22 | 85.7% | C | |
| SkillsBench | 48.2% | 05 | 8 | 42.9% | C | |
| CC-OCR | 81.2% | 06 | 18 | 70.6% | C | |
| C-Eval | 91.4% | 06 | 18 | 70.6% | C | |
| AIME 2026 | 94.1% | 08 | 18 | 58.8% | C | |
| MMMU | 82.9% | 08 | 63 | 88.7% | C | |
| SuperGPQA | 66.0% | 08 | 34 | 78.8% | C | |
| HMMT Feb 26 | 84.3% | 09 | 11 | 20.0% | C | |
| MMLU-Redux | 93.5% | 09 | 48 | 83.0% | C | |
| OCRBench | 89.4% | 09 | 22 | 61.9% | C | |
| RealWorldQA | 84.1% | 09 | 26 | 68.0% | C | |
| VideoMMMU | 84.4% | 09 | 26 | 68.0% | C | |
| Claw-Eval | 60.6% | 10 | 13 | 25.0% | C | |
| ERQA | 62.5% | 10 | 23 | 59.1% | C | |
| HMMT 2025 | 93.8% | 11 | 33 | 68.8% | C | |
| LiveCodeBench v6 | 83.9% | 11 | 53 | 80.8% | C | |
| MMLU-Pro | 86.2% | 12 | 129 | 91.4% | C | |
| SimpleVQA | 0.561 points | 12 | 13 | 8.3% | C | |
| NL2Repo | 36.2% | 13 | 14 | 7.7% | C | |
| IMO-AnswerBench | 80.8% | 15 | 19 | 22.2% | C | |
| SWE-bench Multilingual | 71.3% | 17 | 34 | 51.5% | C | |
| Terminal-Bench 2.0 | 59.3% | 22 | 49 | 56.3% | C | |
| CharXiv-R | 78.4% | 24 | 48 | 51.1% | C | |
| SWE-Bench Verified | 77.2% | 26 | 105 | 76.0% | C | |
| MMMU-Pro | 75.8% | 30 | 66 | 55.4% | C | |
| SWE-Bench Pro | 53.5% | 36 | 45 | 20.4% | C | |
| GPQA | 87.8% | 37 | 234 | 84.5% | C | |
| Humanity's Last Exam | 24.0% | 52 | 93 | 44.6% | C |
Qwen3.6 27B Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
Qwen3.6 27B Runtime Performance
Provider-specific output speed and catalog latency for Qwen3.6-27B. Runtime does not affect the capability score.
| Novita | 7.845 tok/s | 7.801 s | 262.1K | 65.5K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Qwen3.6 27B Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $0.60 input, $3.6 output per 1M
- Official provider
- Alibaba
- Lowest third-party
- From $0.20 input, $1.5 output per 1M via CrofAI
- Tracked offerings
- 26
| QVAC | qwen3.6-27b | global | N/A | N/A | 262.1K | |
| ai& | qwen/qwen3.6-27b | global | N/A | N/A | 262.1K | |
| CrofAI | qwen3.6-27b | global | $0.20 | $1.5 | 262.1K | |
| NanoGPT | alibaba/qwen3.6-27b | global | $0.203 | $2.24 | 260.1K | |
| Merge Gateway | qwen/qwen3.6-27b | global | $0.289 | $2.4 | 131.1K | |
| SiliconFlow | Qwen/Qwen3.6-27B | global | $0.30 | $3.2 | 262.1K | |
| Vultr | Qwen/Qwen3.6-27B | global | $0.30 | $2 | 262.1K | |
| Abacus | Qwen/Qwen3.6-27B | global | $0.32 | $3.2 | 262.1K | |
| Deep Infra | Qwen/Qwen3.6-27B | global | $0.32 | $3.2 | 262.1K | |
| Venice AI | qwen3-6-27b | global | $0.325 | $3.25 | 256K | |
| EmpirioLabs AI | qwen3-6-27b | global | $0.4126 | $2.48 | 256K | |
| Cortecs | qwen3.6-27b | global | $0.446 | $3.01 | 262K | |
| Synthetic | hf:Qwen/Qwen3.6-27B | global | $0.45 | $3.6 | 262.1K | |
| Kilo Gateway | qwen/qwen3.6-27b | global | $0.45 | $2.7 | 262.1K | |
| OVHcloud AI Endpoints | qwen3.6-27b | global | $0.47 | $3.19 | 262.1K | |
| Hugging Face | Qwen/Qwen3.6-27B | global | $0.47 | $3.19 | 262.1K | |
| STACKIT | Qwen/Qwen3.6-27B | global | $0.53 | $0.76 | 262.1K | |
| Regolo AI | qwen3.6-27b | global | $0.58 | $2.42 | 120K | |
| Weights & Biases | Qwen/Qwen3.6-27B | global | $0.60 | $3.6 | 262.1K | |
| Pioneer | Qwen/Qwen3.6-27B | global | $0.60 | $0.60 | 32.8K | |
| Ofox | bailian/qwen3.6-27b | global | $0.60 | $3.6 | 256K | |
| Groq | qwen/qwen3.6-27b | global | $0.60 | $3 | 131.1K | |
| Alibaba | qwen3.6-27b | global | $0.60 | $3.6 | 262.1K | |
| Vercel AI Gateway | alibaba/qwen3.6-27b | global | $0.60 | $3.6 | 256K | |
| OpenRouter | qwen/qwen3.6-27b | global | $0.60 | $3.6 | 262.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Qwen3.6 27B Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Qwen3.6-27B | 60.5 | 27.8B | 262.1K | 65.5K | Yes | Apache 2.0 |
What is Qwen3.6 27B?
Key information about Qwen3.6 27B and its available data.
6 series, supporting both vision-language thinking and non-thinking modes in a single unified checkpoint.
The 64-layer language model uses a hybrid layout of 16 repeats of (3 × Gated DeltaNet → FFN, 1 × Gated Attention → FFN) with hidden dim 5120 and FFN intermediate 17408 — Gated DeltaNet has 48/16 heads for V/QK (head dim 128) and Gated Attention has 24/4 heads for Q/KV (head dim 256).
It supports a native 262,144-token context extensible to ~1,010,000 via YaRN and is trained with multi-token prediction. 8 on GPQA Diamond. 0; accessible via Qwen Studio with the Alibaba Cloud Model Studio API coming soon.
Data as of 2026-08-11.
Qwen3.6 27B vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Qwen3.6 27B
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AC
Qwen3.5 122B A10B
Alibaba Cloud / Qwen Team
AC
Qwen3.5 27B
Alibaba Cloud / Qwen Team
AC
Qwen3.6 35B A3B
Alibaba Cloud / Qwen Team
AC
Qwen3.5 397B A17B
Alibaba Cloud / Qwen Team
AC
Qwen3.6 Plus
Alibaba Cloud / Qwen Team
AC
Qwen3.5 35B A3B
Alibaba Cloud / Qwen Team
FAQ
Common questions about Qwen3.6 27B.
When was Qwen3.6 27B released?
Qwen3.6 27B's default version was released on Apr 21, 2026.
How much does Qwen3.6 27B cost?
Qwen3.6 27B's official API price is $0.60 per million input tokens and $3.6 per million output tokens via Alibaba. The lowest tracked third-party offer starts at $0.20 input and $1.5 output via CrofAI.
Who created Qwen3.6 27B?
Qwen3.6 27B was created by Alibaba Cloud / Qwen Team.
What is the context window for Qwen3.6 27B?
The default version has a 262.1K token context window.
Is Qwen3.6 27B open weight?
Yes. The default version is marked as open weight under Apache 2.0.
How many API providers offer Qwen3.6 27B?
26 provider offerings are linked to the default version.
What models should I compare Qwen3.6 27B with?
Nearby ranked alternatives include GPT-5.1-Instant, Kimi K2 Thinking, LongCat Flash Thinking.