The free, open, on-device language model behind "Smart AI".
Qwen 2.5 0.5B Instruct (Apache 2.0) in Google's MediaPipe format: no account, no key and no cost. It downloads once (about 550 MB) and then runs fully offline on the phone's CPU, which works on any Android phone (arm64, arm32 and x86_64). Its context window is 1,280 tokens, so prompts are kept short.
Properties
- hashCode → int
-
The hash code for this object.
no setterinherited
- runtimeType → Type
-
A representation of the runtime type of the object.
no setterinherited
Methods
-
noSuchMethod(
Invocation invocation) → dynamic -
Invoked when a nonexistent method or property is accessed.
inherited
-
toString(
) → String -
A string representation of this object.
inherited
Operators
-
operator ==(
Object other) → bool -
The equality operator.
inherited
Static Properties
-
downloadProgress
→ ValueNotifier<
int?> -
Download progress 0..100 while a download runs, otherwise null. Shared,
so the progress survives closing and reopening the AI Edit sheet.
final
- supportedOverride ↔ bool?
-
Turns the model off when the device can't run it (tests, desktop).
getter/setter pair
Static Methods
-
download(
) → Future< void> - Downloads the model (once). A second call joins the running download.
-
generate(
String prompt) → Future< String> -
Answers
promptwith the downloaded model. The model is loaded for this call and released afterwards, so it never holds memory while the editor renders video. -
isInstalled(
) → Future< bool> - Whether the model is downloaded and ready to use.
-
remove(
) → Future< void> - Deletes the downloaded model to free space.