Qwen4-27B is in development, based on a GitHub pull request referencing it
1 source, first seen
See it on the rumor trackerRumor
Reddit user finds pull request hinting at Qwen4-27B
A pull request in an Intel-linked vLLM kernels repo references Qwen4-27B and suggests its KV cache size is not shrinking. The author's identity link to Alibaba or Zhejiang Lab is unverified speculation.
Every outlet that ran it 1 article
r/LocalLLaMACommunityLead