v0.32.1

playcanvas/enginev0.32.1Jul 16, 2026by github-actions[bot]

AI Summary

This update focuses on enhancing AI agent capabilities, specifically improving tool calling for Gemma 4 and fixing memory leaks in the MLX model cache.

Key Highlights

Improved Gemma 4 tool calling and multi-turn reasoning reliability
Fixed MLX model cache leak to prevent memory usage increase
MLX text model loading now respects `OLLAMA_LOAD_TIMEOUT`
Interactive agent now receives current working directory for better context
Fixed `ollama launch` to properly open model picker for deprecated models

New Features

Enhanced Gemma 4 tool calling
MLX cache performance improvements
Agent context awareness (CWD)
Sign-in prompts for auth-required operations

Full Release Notes

## What's Changed
- Improved Gemma 4 tool calling and multi-turn reasoning, including more reliable tool-response continuations
- Fixed a recurrent MLX model cache leak that could increase memory use across requests, and improved cache snapshot performance
- MLX text model loading now respects `OLLAMA_LOAD_TIMEOUT`
- Agent web search and fetch now tell users to run `ollama signin` when authentication is required
- The interactive agent now receives the current working directory for better project context
- Fixed `ollama launch` so choosing **Pick another model** for a deprecated model passed with `--model` opens the model picker
- Updated VS Code setup documentation for the official Ollama extension

**Full Changelog**: https://github.com/ollama/ollama/compare/v0.32.0...v0.32.1-rc0