Developers · Updated 21 Oct, 02:45 am IST
Ollama v0.40.7-rc1 deduplicates CUDA runtime payloads in MLX and GGML backends
Why it matters for readers: It's a maintenance release that tightens how Ollama bundles and verifies CUDA and math libraries.
- Shared CUDA libraries for GGML and MLX were moved into a sibling CUDA payload to avoid duplication.1
- In-process and subprocess MLX loading were updated to resolve the packaged dependency directory for the CUDA payload.1
- MLX-private libraries and JIT headers remain in the MLX overlay while the full OpenBLAS Fortran runtime closure is included.1
- The release packaging now verifies those dependencies as part of the v0.40.7-rc1 assets.1
Get a brief like this every morning
Uzha reads hundreds of sources and gives you the stories that matter for your work, with every source linked. Free.
Get started