The surprising part: Qwen 3.8 Flash Next is being tried locally in DwarfStar on Apple Silicon, with a claimed 65 t/s on an M3 Ultra.
The test ask is specific: 64GB Mac, Q2 weights on HF, PR branch on GitHub.
When you run out, spin up Qwen 3.8 Flash Next on the DGX !!
And Qwen 3.8 Flash Next NVFP4 on one DGX operates my OBSBOT_Official and scans the room.
Today's Qwen 3.8 Flash Next NVFP4 TP=1 DGX Spark update includes native 4096×4096 image processing
Multi-image history, xhigh thinking by default, reasoning retention, and a 4 GiB shared-memory image cache. Three hours of sustained agentic OpenCode work, including visual tasks
You have reached the end of the archive
All of qwen38