Qwen3.6-27BOpen Weight

Released

Qwen3.6-27B is an AI model released by Qwen on Wednesday, Apr 22 2026, 6 days after Qwen3.6. It is an open-weight model — the trained weights are available to download and run. It is a 27B parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model with a 256k token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover SWE-Bench Verified, SWE-Bench Multilingual, NL2Repo-Bench, LiveCodeBench, Terminal-Bench 2.0, Humanity's Last Exam, and 6 more.

Benchmarks

Coding

SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better.
77.2%
#25 of 95
SWE-Bench MultilingualMultilingual coding — Like SWE-Bench, but the coding problems span many programming languages, not just one. Tests how broadly the AI can code. Higher is better.
71.3%
#21 of 43
NL2Repo-BenchRepo-level code generation — Tests whether the AI can turn a natural-language requirement into working code across an entire repository, not just produce a single function or patch. Higher is better.
36.2%
#12 of 13
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better.
83.9%
#15 of 86

Terminal & CLI

Terminal-Bench 2.0Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? (Version 2.0 of the test.) Higher is better.
59.3%
#16 of 37

Reasoning & science

Humanity's Last ExamMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better.
24%no tools
#44 of 58
AIME 2026Reasoning — single-answer accuracy on the 2026 invitational mathematics exam without tools
94.1%
#4 of 6
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
87.8%
#38 of 145

Knowledge work

MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark
86.2%
#5 of 58

Multimodal

CharXiv ReasoningChart reasoning — Can the AI read and reason about complex charts and figures, not just text? Higher is better.
78.4%
#22 of 32
MMMU-ProMultimodal reasoning — A tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better.
75.8%
#22 of 36
MMMUMultimodal — Tests the AI on understanding images and text together across many college subjects. Higher is better.
82.9%
#5 of 50

Compare Qwen3.6-27B with

Qwen3.6-27B

About

Qwen3.6-27B was released on April 22, 2026 as a dense open-weight alternative to the 35B-A3B mixture-of-experts model. It retained text and visual inputs and both thinking and non-thinking modes.

In Qwen's launch evaluation, the 27B checkpoint scored 77.2 on SWE-bench Verified and 59.3 on Terminal-Bench 2.0. The dense architecture gave developers a smaller total weight count with a different inference profile from the sparse model.

Frequently asked questions

Qwen3.6-27B was released by Qwen on Wednesday, Apr 22 2026.

All Qwen releases

42 tracked