🚀 RELEASES & LAUNCHES
- Qwen4 code PR hints at 1M-token context window; family announced at Apsara
A Qwen-code PR suggests Qwen4 models will have a 1M-token context window, following Alibaba's September 22 announcement of the Qwen4 family.- For local-AI builders and Qwen developers tracking the next open-weights release.
- A pull request in QwenLM/qwen-code references a 1 million-token context window for Qwen4 models.
- Alibaba announced the Qwen4 family (Max, Plus, Flash, and an open-weights 27B) at Apsara on September 22, saying Qwen4 is in training and coming 'very soon'.
- No specs, license, or release date are confirmed; community speculation points to late October or November.
- The Qwen3.8-Flash-Next preview (125B total / 6B active MoE with an n-gram memory table) is the official architecture preview for Qwen4.