●Change8

v0.18.2-rc0

📦 ollamaView on GitHub →
✨ 3 features🐛 1 fixes

Summary

This release introduces significant performance and feature enhancements for the MLX backend, including model eviction, quantized embeddings, and fast SwiGLU. It also includes a fix for the web_search legacy path in the cloud proxy.

Migration Steps

  1. Register websearch for openclaw environments.

✨ New Features

  • Model eviction implemented for MLX backend.
  • Added prequantized tensor packing and related changes for qwen35 support in MLX.
  • Implemented quantized embeddings and fast SwiGLU activation function for MLX, along with runtime fixes.

🐛 Bug Fixes

  • Cloud proxy now flushes on newlines for the web_search legacy path.