Completes or generates text from a prompt. Similar to using a local LLM (Ollama style).
The model will be downloaded from Hugging Face Hub (~967 MB) and cached in your browser (IndexedDB). Subsequent runs won't need to download again. All processing happens locally — no data ever leaves your device.
Connects to huggingface.co · *.cdn.hf.co · cdn.jsdelivr.net — only to download the file: nothing of yours is sent.