Support

How can we help?

Common questions

Which model should I start with?

Start with the built-in SmolLM2 360M download (about 386 MB). For better answers, open Models, then Browse Hugging Face, and pick a file marked "Fits comfortably". On most recent iPhones, 1 to 3 billion parameter models at Q4_K_M quantization are a good balance of quality and speed.

A model fails to load or the app closes while loading.

The model is probably too large for your phone's memory. Try a smaller model or a smaller quantization (for example Q4_K_M instead of Q8_0), and close other apps first.

Why does my phone get warm?

Running a language model uses the GPU heavily. The Lab's endurance test is designed to measure exactly this, and it stops automatically if the phone reaches a critical thermal state or the battery runs low while unplugged.

Does it work offline?

Yes. Once a model is downloaded, chatting and every lab feature work with no connection. Only browsing and downloading models, and the optional remote provider feature, need the internet.

How do I delete my data?

Clear the lab notebook in Settings, delete individual sessions from the home screen, and remove model files on the Models screen. Deleting the app removes everything. See the privacy policy.

Is the model always right?

No. Small on-device models make mistakes and can state wrong things confidently. The confidence view shows how sure the model was of each word, but it does not make answers correct. Check anything important.