AI Navigate

v0.18.2

Ollama Releases / 3/19/2026

📰 NewsDeveloper Stack & InfrastructureTools & Practical Usage

Key Points

  • v0.18.2 adds model eviction for MLX to improve model lifecycle management, as implemented by @jessegross in PR #14569.
  • mlx adds prequantized tensor packing and changes for qwen35, contributed by @pdevine in PR #14878.
  • mlx introduces quantized embeddings, faster SwiGLU, and runtime fixes, also by @pdevine in PR #14884.
  • cloud_proxy improves the web_search legacy path by flushing on newline, implemented by @drifkin in PR #14897.
  • launch registers websearch for OpenClaw, contributed by @ParthSareen.

What's Changed

Full Changelog: v0.18.1...v0.18.2-rc1