Quick Answer
Check ollama.com or the Ollama GitHub releases page for the current version. On Linux, run the install script to update. On Mac/Windows, download the latest installer.
Updated: 2026-05
Key Takeaways
As of May 2026, the latest stable Ollama release is listed at github.com/ollama/ollama/releases — that page shows the exact version number and release notes. This page is updated monthly — last verified May 2026.
Run ollama --version in a terminal to see your installed build. The output is one line. If the GitHub releases page shows a higher version number, your install is behind.
New model families — including Llama 3.1 and Qwen 3 — require a minimum Ollama version to load correctly. Recent releases also include faster GGUF loading and better multi-GPU context handling — improvements that compound across daily use.
Updating is safe: your downloaded models are stored in a separate directory and are never removed by an Ollama update. The user-facing REST API has maintained backwards compatibility since version 0.1.
If you are several versions behind, run ollama list before updating to note your current models. Nothing will break, but this 30-second step eliminates uncertainty before a major version jump.
For a complete changelog of what each Ollama release changed and which model families benefit, see the local LLM model updates guide for 2026.
| Platform | Update Method | Notes |
|---|---|---|
| Linux | Run install script (one command) | Overwrites binary, keeps models |
| macOS | Re-download from ollama.com | Drag to Applications, models kept |
| Windows | Re-download .exe from ollama.com | Run installer, models kept |
curl -fsSL https://ollama.com/install.sh | shollama --version in a terminal. It prints your build number in one line, for example ollama version 0.x.y.~/.ollama/models/ on Linux and macOS, separate from the application binary. Updates never touch that directory. See the best Ollama models right now if you want to add new ones after updating.