Starstruk Logo MarkStarstruk

v0.30.8

OpenMOSS/MOSS-TTSv0.30.8Jun 12, 2026by github-actions[bot]

AI Summary

This release focuses on enhancing the stability and performance of MLX inference. It includes critical bug fixes for provider selection, improvements to prompt caching for better KV reuse, and new reliability features for speculative decoding.

Key Highlights

  • Fixed `ollama launch` selecting the wrong provider in some cases
  • Improved prompt caching by decoupling it from context shift for better KV cache reuse
  • More stable MLX inference with hardened linear and embedding layers
  • MLX runner now creates snapshots during prompt processing and speculative decoding
  • Improved recurrent model support with per-boundary states from the gated-delta kernels

New Features

  • Enhanced prompt caching mechanism
  • Improved recurrent model support
  • MLX runner reliability via snapshots
  • Hardened linear and embedding layers

Full Release Notes

## What's Changed
* Fixed `ollama launch` selecting the wrong provider in some cases
* Improved prompt caching by decoupling it from context shift for better KV cache reuse
* More stable MLX inference with hardened linear and embedding layers
* MLX runner now creates snapshots during prompt processing and speculative decoding for improved reliability
* Improved recurrent model support with per-boundary states from the gated-delta kernels

**Full Changelog**: https://github.com/ollama/ollama/compare/v0.30.7...v0.30.8