Qwen3.6-27B Released: Dense Multimodal Model Beating Qwen3.5 Flagship

After the releases of Qwen3.6-Plus and Qwen3.6-35B-A3B, Qwen3.6-27B has been released — a dense multimodal model with 27B parameters for developers, surpassing the previous flagship Qwen3.5-397B-A17B (397B)

The model is ~15 times smaller than its predecessor flagship, yet it outperforms it on SWE-bench, Terminal-Bench, and other benchmarks

Works in openclaw, qwen code, and claude code via a compatible API

Code to run in Claude Code
export ANTHROPIC_MODEL="qwen3.6-27b"
export ANTHROPIC_BASE_URL=https://dashscope-intl.aliyuncs.com/apps/anthropic
export ANTHROPIC_AUTH_TOKEN=
claude

Qwen Studio
HuggingFace (default) | HuggingFace (FP8)
ModelScope