Alibaba
Qwen3.6 27B (Reasoning)
The Qwen3.6 35B-A3B native vision-language model is built on a hybrid architecture that integrates linear attention mechanisms with a sparse mixture-of-experts framework, achieving higher inference efficiency. Compared with the 3.5-35B-A3B, this model demonstrates significantly improved agentic coding capabilities, mathematical and code reasoning abilities, spatial intelligence, as well as object localization and object detection performance.
Qwen model releasesModel Snapshot
TypeLanguage
ReleaseApr 22, 2026
Context256,000
Tagsreasoning, tool-use, implicit-caching, file-input
Overall
Rank #37 across 526
45.8
Coding
Rank #81 across 436
36.5
Blended Price
Rank #227 across 357
$1.35/M
Strength Profile
Percentile score by analysis domain.
* Cost is inverted: lower input, output, and blended prices rank higher.
Benchmark Percentiles
Higher bars mean stronger relative placement.
Model Details
Runtime
Output Speed54.7 tok/s
First Token1.55s
Max Output256,000
Pricing
Blended$1.35/M
Input$0.600/M
All Benchmarks
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 45.8 | #37 |
| Artificial Analysis Coding Index | coding | 36.5 | #81 |
| GPQA | reasoning | 84.2% | #75 |
| Humanity's Last Exam | reasoning | 21.6% | #72 |
| SciCode |