Model Library
Pre-quantized models ready for immediate deployment. Mirai automatically handles format conversion (CoreML, ONNX, TFLite).
Text4.7GB
Llama 3 8B
Most popular general purpose model. Great for chat.
#Chat#Reasoning
2.4MText4.1GB
Mistral 7B v0.3
High performance instruction tuned model.
#Code#Chat
1.8MText5.2GB
Gemma 2 9B
Google's open weights model optimized for TPUs.
#Reasoning
900KAudio1.5GB
Whisper Large v3
Robust speech recognition supporting 99 languages.
#ASR#Multilingual
3.1MImage6.8GB
Stable Diffusion 3
Text-to-image generation optimized for Apple Neural Engine.
#Generation#Art
450KText2.3GB
Phi-3 Mini
Efficient small language model for mobile devices.
#Mobile#Fast
1.2M