Models
Aevori chooses the best verified local option for your device.
Built-in
Apple’s on-device system language model is preferred when the device, Apple Intelligence setting, language, and runtime availability permit it.
Lite
Qwen3.5 0.8B MLX 4-bit. Repository snapshot: about 0.61 GiB.
Balanced
Qwen3.5 2B MLX 4-bit. Repository snapshot: about 1.63 GiB. This is the provisional downloadable default pending physical-device quality and thermal tests.
High quality
Qwen3.5 4B MLX 4-bit. About 2.85 GiB and available only after device benchmarks approve it.
Actual installed size can differ. Aevori checks free space and model integrity before use.