Ollama v0.32.1–v0.32.2: Gemma 4 Tool Calling, Claude Code Channel Fix, MLX Memory Leak Fix

What's new

• Gemma 4 tool calling (v0.32.1): improved multi-turn reasoning and tool calling for Gemma 4 models
• Claude Code channels (v0.32.2): fixes regression from v0.32.0 — Claude Code channels are available again
• Unlimited tool-calling rounds (v0.32.2): cloud models no longer have a limit on tool-calling rounds
• MLX memory leak fix (v0.32.1): fixes memory leak in MLX model cache that increased memory consumption with repeated use
• Timeout respect: text model loading now respects timeout settings
• Agent working directory: interactive agent receives current working directory for project context
• GCC 13 toolchain (v0.32.2)

Why it matters

The Claude Code channel fix is critical for developers combining Ollama with Claude Code. The MLX memory leak fix matters for Mac users with long sessions. Unlimited tool-calling rounds is important for agentic workflows.

How to try it

Run ollama update or download from github.com/ollama/ollama/releases. Recommended version: v0.32.2.