FRIDAY supports multiple Llama 3.2 variants — all running locally on your iPhone. Pick the right balance of speed and intelligence for your device.
The lightest model — blazing fast responses with minimal RAM usage. Best for quick questions, short tasks, and older devices where speed matters more than depth.
The default FRIDAY model — the best balance of intelligence and speed on modern iPhones. Handles complex reasoning, code, writing, and nuanced conversations with ease.
Maximum on-device capability — deep reasoning, long-form content, complex coding and analysis. Requires an iPhone 16 Pro or Apple silicon iPad. Coming in a future update.
| Capability | 1B — Fastest | 3B — Recommended | 8B — Coming soon |
|---|---|---|---|
| Runs on-device | ✓ | ✓ | ✓ |
| Works offline | ✓ | ✓ | ✓ |
| Zero data sent | ✓ | ✓ | ✓ |
| Conversational Q&A | ✓ | ✓ | ✓ |
| Code generation | Basic | Strong | Excellent |
| Long-form writing | Short only | Good | Excellent |
| Multi-step reasoning | Limited | Good | Very good |
| Minimum iPhone | iPhone 12 | iPhone 14 | iPhone 16 Pro |
| Available now | ✓ | ✓ | Soon |
For tasks that need more than any on-device model can deliver today, FRIDAY lets you connect your own API key to access the world's most capable cloud models. Your key goes directly to the provider — FRIDAY is just the interface.
Connect your OpenAI API key to access GPT-4o — one of the most capable language models available. Ideal for complex reasoning, long documents, and advanced coding tasks.
Use your Anthropic API key to access Claude Sonnet — excellent at nuanced writing, analysis, and thoughtful multi-step reasoning with a very long context window.
Cloud features are fully optional. The default on-device experience never requires an API key or internet connection.
Performance depends on RAM and Neural Engine generation. More RAM = bigger model = smarter responses.
4 GB RAM · A14 / A15
6 GB RAM · A15 / A16
8 GB RAM · A17 / A18
8 GB RAM · A18 Pro