Qwen 3.8 27B model reaches one million downloads

The Qwen 3.8 27B open-weights model has achieved significant traction in the developer community, surpassing one million downloads within two days of its release and appearing on Hugging Face’s global trend list.

According to Artificial Analysis, the model holds a score of 52 on its Intelligence Index, placing it in a high-performing class alongside models like DeepSeek V4 Flash. Due to its ability to perform complex reasoning tasks while remaining small enough to run on consumer-grade hardware—such as a 24GB GPU via 4-bit quantization—it has been nicknamed ‘local Opus 4.6’ by some developers.

Testing has demonstrated the model's advanced capabilities in specialized tasks, including code reasoning and reverse-engineering. It has also become a preferred local model for users of the Cline coding agent within days of its debut.