Qwen3.8-27B
Points and comments are a snapshot, not live.
Alibaba releases open-weight Qwen3.8-27B model outperforming Qwen3.7-Plus.
Qwen (@Alibaba_Qwen) released open weights for Qwen3.8-27B, a native multimodal dense model with 27B parameters. It outperforms Qwen3.7-Plus overall, excelling in coding and office workflows. The model has 262K native context, extendable to 1M via YaRN, and is licensed under Apache 2.0. Open weights for the larger Qwen3.8-2.4T-A95B (Max-level) were also released. Available on Hugging Face and ModelScope. Unsloth AI created Dynamic GGUFs for local runtime and support in Unsloth Desktop.
What commenters are saying
Comment #3 reports that Qwen3.6-27B has been the main LLM powering their agentic stack, using a test-and-verify approach for models. They are excited to test the new 3.8 version against Qwen3.6. Two other comments point to separate threads and where discussion has moved.