Hugging Face models and easier downloads

You can now download compatible GGUF models from Hugging Face through your Ollama connection. Paste a model link, an Ollama model name, or an ollama run or ollama pull command into the new Download Model dialog. For Hugging Face models, leave the variant on Automatic or choose a quantization or an exact GGUF filename.

• Paste a model link into model search to open a download result. The dialog shows the model source, reference and destination before you start.
• Choose New Chat from a completed download notification to start a conversation with that model selected. Already installed chat models offer the same action directly in the download dialog.
• Open an installed Hugging Face model’s repository from its model row, or search for Hugging Face to find models from that source.
• Failed downloads offer Edit and Retry, with clearer guidance about repository access, model variants and Ollama compatibility. Error details remain available in Download Jobs.
• The progress panel clears when a download completes, fails or is cancelled. Cancelled downloads also refresh the model list, and progress from a stopped download no longer interferes with a later one.
• Connect a Provider during first-time setup now opens the provider chooser, so you can choose managed Ollama, an existing server or a cloud provider.
• Chat bubble backgrounds display correctly with custom GTK themes.
• Model search fits better in narrow windows, and installed model names have more room to wrap.

Read about Hugging Face model downloads, retrying a failed download, or getting started.