Rapid-MLXProduct update
Rapid-MLX · v0.14.0
Rapid-MLX v0.14.0 adds experimental Computer Use, which once enabled under Experimental Features plans and runs bounded tasks locally across supported Mac apps, starting with Downloads cleanup and drafting content to post through Safari or Chrome. A new experimental Share Compute workspace lets users who opt in offer local capacity to a configured compute pool, and it stays off by default. On an M3 Ultra the project measured Qwen3.8 27B MTP verification decode rising from 45.87 to 52.28 tok/s, with some acceleration paths remaining opt-in. The catalog adds text models such as Granite 4.2 and MiniCPM5 2B and image models including FLUX.1 schnell and SDXL Base.
Saved . A shortened excerpt from GitHub Releases; the original publisher retains authorship.