Compare Models

Find the right model for your phone.

All models in Orus run 100% on your device hardware with zero server uploads and zero internet required. Easily switch or download models in Settings > Models.

Qwen 2.5 0.5B

Ultra-compact and lightning fast for every Android device.

469 MBStorage on phone
Up to 45 t/sProcessing speed
4 GB minimumRequired memory
—Voice & Hearing

Instant queries, translation, quick notes, and all-day battery preservation.

Recommended for

Llama 3.2 1B

Meta's edge architecture tuned for concise, structured reasoning.

770 MBStorage on phone
Up to 32 t/sProcessing speed
6 GB minimumRequired memory
—Voice & Hearing

Summarizing long articles, drafting emails, and quick structured thinking.

Recommended for

Qwen 2.5 1.5B

The optimal everyday companion with hands-free voice interaction.

1.0 GBStorage on phone
Up to 26 t/sProcessing speed
6 GB minimumRequired memory
Full duplex voice & hearingVoice & Hearing

Natural conversations, voice assistant tasks, and general daily intelligence.

Recommended for

Llama 3.2 3B

Frontier capabilities for complex problem solving and deeper nuance.

2.0 GBStorage on phone
Up to 18 t/sProcessing speed
8 GB minimumRequired memory
Full duplex voice & hearingVoice & Hearing

Heavy academic study, complex creative writing, and advanced programming.

Recommended for

Custom Models

Import any model from Hugging Face.

You are never locked into pre-configured options. In Settings > Models, tap Add a model to download any public GGUF weights directly from Hugging Face or import existing checkpoint files from your phone storage.

Orus automatically detects available RAM and processor threads to optimize context windows and thermal performance for imported weights.

Orus Models Settings on Android

Experience private AI on your phone.

Download Orus directly as an APK. Zero accounts, zero cloud tracking, and completely offline from the first launch.