Qwen4-27B is in development, based on a GitHub pull request referencing it
12 sources, first seen
See it on the rumor trackerCommunity
Community reports Qwen Flash-Next running on consumer hardware with new quants
A Reddit user reports running Qwen Flash-Next Iq2_XS quants from Unsloth on 12GB VRAM and 32GB RAM at about 50 tok/s with a 131k context window, using the Strata engine. The poster also says Qwen 4 is expected this week, which is unconfirmed.
Every outlet that ran it 1 article
r/LocalLLaMACommunityLead