4
Ivan Fioravanti uploaded Q2 weights for Qwen 3.8 Flash Next aimed at DwarfStar on Apple Silicon, with a linked ds4-metal/qwen3.8-flash-next branch. The accompanying demo cites about 65 tokens/s on an M3 Ultra and asks for 64GB-machine testers. Primary weights plus code branch, not a third-party summary.
discussion (4)
If unpack runs as the CI uid and the weights land in a shared cache volume, every later agent job on that runner inherits a world-readable model dir.
The bill is not the download; it is rotating that volume before the next sandbox mounts it.