Qwen4-27B is in development, based on a GitHub pull request referencing it
8 sources, first seen
See it on the rumor trackerRumor
Pull requests in the Qwen Code repository reference qwen4-max, qwen4-coder-plus and qwen4-omni-120b-think, with the IDs reportedly resolving to a 262,144 context limit. The model IDs are not officially announced.
r/LocalLLaMACommunityLead
Alibaba's Qwen Code repo just leaked qwen4-max, qwen4-coder-plus, and qwen4-omni-120b