Alibaba releases Qwen3.6-27B, beating larger predecessor on coding benchmarks

Alibaba has released Qwen3.6-27B, a new dense open-source model with 27 billion parameters that outperforms its much larger predecessor, Qwen3.5-397B-A17B, on nearly every coding benchmark tested.

It scored 77.2 on SWE-bench Verified compared to the predecessor's 76.2.
It achieved 59.3 on Terminal-Bench 2.0 versus 52.5 for Qwen3.5-397B-A17B.
The model handles both text and multimodal reasoning tasks.
As a dense model, it is easier to run than complex Mixture of Experts architectures.

The model is available through Qwen Studio, the Alibaba Cloud Model Studio API, and as open weights on Hugging Face and ModelScope, aiming at developers who want strong coding performance without dealing with a massive model.

Benchmarks

Benchmark	Model	Score
SWE-bench Verified	Qwen3.6-27B	77.2pts
Terminal-Bench	Qwen3.6-27B	59.3pts

Benchmark

Model

Score

SWE-bench Verified

Qwen3.6-27B

77.2pts

Terminal-Bench

Qwen3.6-27B

59.3pts