Ollama v0.40.5-rc1 adds Nemotron 3.5 support and MLX improvements

Brief written by the FOSSWire AI editor from Ollama How we write

October 15, 2026 1 min read 32 reads

Key Takeaway

Ollama's latest release candidate introduces Nemotron 3.5 support and significant bug fixes for the MLX backend.

Summary

Ollama has released version 0.40.5-rc1, which includes support for the Nemotron 3.5 model. The update focuses on improving the MLX backend, specifically addressing Nemotron bugs, stabilizing Mamba components, and enhancing global scale support for various quantization formats.

MLX and Windows build updates

MLX Backend: Added support for Nemotron 3.5 and fixed layer-count inference and MTP quantization issues.
Scaling Support: Added support for per-expert F32 global scales and preserved activation-scale companions in ModelOpt and compressed-tensors NVFP4 imports.
Mamba Improvements: Stabilized Mamba softplus and compiled Mamba gate and norm scaling.
Windows CI: Shortened Windows Vulkan build paths to prevent exceeding Windows path length limits during shader-generator compiler probes.

Original reference: github.com →