v0.22.1
Local LLM runner gets smarter Gemma 4 reasoning and seamless model recommendations.
Ollama, the popular open-source tool for running large language models locally, has shipped version 0.22.1. This release primarily targets the Gemma 4 model renderer, improving both its reasoning ('thinking') capabilities and its ability to perform tool calling. Tool calling allows the model to interact with external functions, APIs, or code execution, making Gemma 4 more practical for agent-like workflows on local hardware. The update is relatively lightweight — it includes only a handful of changes since v0.22.0, with 3 commits to the main branch.
Beyond the Gemma 4 improvements, Ollama v0.22.1 introduces a quality-of-life enhancement: model recommendations are now updated dynamically without requiring users to upgrade the Ollama client itself. This reduces friction for those who rely on curated model lists. The desktop application’s launch page was also aligned with the `ollama launch` CLI integrations, ensuring a consistent experience across interfaces. A minor fix corrected the title for the Poolside integration. The release is signed with a verified GPG key and includes 18 asset files for various platforms.
- Gemma 4 renderer updated to improve thinking (reasoning) and tool calling abilities
- Model recommendations now update without needing to upgrade the Ollama client
- Desktop app launch page aligned with 'ollama launch' CLI integrations; Poolside integration title fixed
Why It Matters
Better local reasoning and tool use for Gemma 4, plus smoother model discovery — no full update required.