Release Overview

Ollama has released v0.17.3, a minor patch addressing a specific parsing issue with the Qwen 3 and Qwen 3.5 model families.

What's Fixed

Tool Call Parsing During Reasoning: The release fixes a bug where tool calls (function invocations) in Qwen 3 and Qwen 3.5 models would fail to parse correctly when emitted during the model's internal thinking phase. This is particularly important for users leveraging these models' reasoning capabilities with function calling features.

Impact for Developers

If you're using Qwen 3 or Qwen 3.5 models with tool use (also known as function calling), you should update to v0.17.3 to ensure reliable tool invocation parsing. This fix ensures that agent workflows, multi-step reasoning tasks, and structured tool use work as expected.