AI Models
Download, manage, and switch between GGUF models that run entirely on-device via llama.cpp.
Available Models
| Model | Download size | Min RAM | Capabilities |
|---|---|---|---|
| Qwen3.5 0.8B | 0.51 GB | 2 GB | Text only — fastest, most limited context |
| Qwen3.5 2B | 1.2 GB | 3 GB | Text only — balanced speed and quality |
| Qwen3.5 4B | 2.5 GB | 5 GB | Text only — best quality on-device |
| Qwen3.5 4B Vision | 2.5 GB + mmproj | 5 GB | Text + photo analysis — single download, no separate vision model needed |
All four models support Thinking mode (Qwen3's <think>... reasoning blocks).
Recommended Model
The app recommends a model automatically based on your device's physical RAM:
| Device RAM | Recommended model | Why |
|---|---|---|
| 2–3 GB | Qwen3.5 0.8B | Only model that fits |
| 3–5 GB | Qwen3.5 2B | Best balance of speed and quality |
| 5 GB+ | Qwen3.5 4B | Full intelligence and longest context |
Models exceeding your device's RAM are hidden from the selection list.
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 a different 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 required. Downloads are blocked on cellular to protect your data plan.
Background downloads
LucidPal uses iOS background transfer sessions, so the download continues even when the app is suspended or the screen is locked. If the app is terminated while a download is in progress, iOS resumes the transfer automatically when you reopen the app.
Resuming interrupted downloads
If your WiFi drops mid-download, LucidPal saves resume data automatically. The next time you tap Retry, the download picks up from where it stopped — it does not restart from 0%. If the resume data becomes corrupt, the app detects this and restarts cleanly.
Integrity check
Every download is verified against a SHA-256 checksum sourced from HuggingFace LFS metadata. If the file is corrupt, LucidPal deletes it and retries automatically (up to 2 times).
Integrated Vision Models
Qwen3.5 4B Vision is an integrated model — it handles both text and images in a single download. When you select it:
- One file covers everything: text chat, calendar operations, and image understanding
- No separate vision model download required
- Vision features are available immediately after the single download completes
When you select an integrated model in onboarding or in Settings → Text Model, an Integrated badge appears next to its name. The app automatically hides text-only models of the same tier to avoid confusion.
Switching Models
Go to Settings → Text Model to switch between downloaded models. The new model loads the next time you start a conversation.
Vision Model (Separate Download)
If you use Qwen3.5 4B (text-only) as your main model, you can separately download the vision model in Settings → Vision Model. This lets you run the faster 4B text model and switch to vision only when needed.
The vision model requires:
- A separate mmproj file (vision projector, ~1.5 GB)
- The main vision GGUF file (~2.5 GB)
Both are downloaded together from the same screen.
Context Window
Context window size is determined by your device's RAM:
| Device RAM | Max context | History kept |
|---|---|---|
| < 6 GB | 4K tokens | ~20 messages |
| ≥ 6 GB | 8K tokens | ~50 messages |
You can reduce the context window in Settings → Inference → Context Size for more RAM headroom on very large models.
Custom Models
Custom GGUF model support is planned for a future release.
Deleting a Model
- Go to Settings → Text Model (or Vision Model)
- Tap the model you want to remove
- Tap Delete Model
The file is removed immediately. You can re-download it at any time.
Storage
Models are stored in the app's Documents folder accessible via the Files app under On My iPhone → LucidPal.
Deleting the app also removes all downloaded model files.