Selected model
Qwen 3.5 0.8B
Fastest path. Best chance of working on modern integrated GPUs and Apple Silicon.
Quick demos, weaker hardware, first-time testing~1-2 GB VRAM recommended
New chat
Start a local conversation.
Choose a model, tune the inference settings, and begin writing below. Everything stays in your browser.
0 charactersIdle