AI Models
Choosing and downloading the right model for your iPhone.
LucidPal runs an AI model entirely on your device. On first launch, you'll be prompted to download one. The app recommends the best model for your iPhone automatically.
Available Models
| Model | Download size | Min RAM | Best for |
|---|---|---|---|
| Qwen3.5 0.8B | 0.51 GB | 2 GB RAM | Older iPhones with limited RAM |
| Qwen3.5 2B | 1.2 GB | 3 GB RAM | iPhone 12 / 13 — recommended default |
| Qwen3.5 4B | 2.5 GB | 5 GB RAM | iPhone 14 Pro / 15 / 16 and newer |
All three models run at similar speed relative to their size — the 4B model gives more accurate and nuanced responses.
Downloading a Model
- Open LucidPal for the first time — the Model Download screen appears automatically.
- Check the recommended model — the app selects the best model for your device RAM. You can switch to the other model if you prefer.
- Tap Download — the download runs in the background. You can lock your screen — it continues automatically.
- Start chatting — once the download completes, LucidPal loads the model and you're ready to go.
Wi-Fi is strongly recommended. The models are between 0.51 GB and 2.5 GB.
Switching Models
Go to Settings → AI Model to switch between models. The new model loads the next time you start a conversation.
Storage
Models are stored in the app's local storage on your device. Deleting the app also removes the downloaded model file.
Thinking Mode
Qwen3.5 models support a Thinking mode where the AI reasons through your request before answering. This improves accuracy for complex calendar operations.
Toggle it in Settings → Thinking Mode. When enabled, you can tap the Thinking disclosure in any assistant reply to see the reasoning steps.
Thinking mode uses slightly more processing time but produces better results for multi-step calendar requests like:
"Schedule a 2-hour block every Tuesday this month, but skip the week of March 10th"