Every feature runs on your device. No internet required. No subscription. No compromise on privacy.
FRIDAY runs Llama 3.2 entirely on your iPhone using Apple's Neural Engine and Metal GPU acceleration. Responses generate at 20–40 tokens per second — no servers, no latency spikes, no downtime.
Your messages never leave your phone. Every word you type stays on-device, encrypted by iOS, never logged.
Full conversation history with context window management. FRIDAY remembers the whole thread while you work.
All your conversations are saved locally. Browse, search, and resume past threads anytime — even offline.
Need GPT-4o or Claude Sonnet for a demanding task? FRIDAY lets you connect your own API key for cloud inference — your key, your cost, zero middlemen. Switch back to on-device instantly. Cloud is always opt-in.
Dictate your messages with on-device speech recognition. Transcription happens locally — nothing sent to Apple or anyone else.
Once the model downloads, FRIDAY works anywhere — no signal required. Flights, remote areas, wherever you go.
No round-trip to a cloud server. Inference starts immediately on your hardware. No waiting for API rate limits or outages.
Code blocks, bold, bullets, headers — FRIDAY renders rich text responses so output is always readable and well-structured.
When you download FRIDAY, the Llama 3.2 3B model is stored directly on your device. Every token is generated by your iPhone's Neural Engine — not a server farm in Virginia.
This means no API limits, no latency from geographic routing, no service outages. Your AI is always available because it's always with you.
Explore models →
For tasks that need heavier reasoning — complex code, long documents, multi-step plans — FRIDAY lets you attach your own OpenAI or Anthropic API key.
Your key goes directly from your phone to the provider. FRIDAY never sees it, stores it, or proxies it through any server. Keychain-encrypted on your device.
When you're done, switch back to on-device with one tap. The cloud is a tool, not a requirement.
FRIDAY supports the two most capable AI providers on the market as optional backends. Both require your own API key — FRIDAY acts as a thin local client, never a proxy.
Every design decision in FRIDAY starts with the same question: does this require sending data anywhere? If yes, it either doesn't ship or it's optional.