Conversion and optimization toolkit
Convert and optimize your model for iPhone, iPad and Mac.
Convert and optimize your model for Apple devices.
One command to get your model running on 2B Apple devices.
One command to get your model running on 2 billion Apple devices.
Step 1
Convert in one command.
If your model is based on one of popular architectures, including your own fine-tunes and adaptations, it converts in one command.
Gemma
Polaris
HuggingFace
DeepSeek
Llama
Qwen
Hand-written converter for each model family.
Fine-tunes and adaptations supported out of the box.
Converts from Hugging Face or any supported source.
Step 1
Convert in
one command.
If your model is based on one of popular architectures, including your own fine-tunes and adaptations, it converts in one command.
Gemma
Polaris
HuggingFace
DeepSeek
Llama
Qwen
Hand-written converter for each model family.
Fine-tunes and adaptations supported out of the box.
Converts from Hugging Face or any supported source.
Step 2
Run on any Apple device.
Your converted model runs on key Apple defices through Mirai's inference engine. Same model, same output, every device.
Runs on iPhone, iPad, and Mac.
Full performance metrics on every run.
Same model, every device. Convert once.
Step 2
Run on any Apple device.
Your converted model runs on key Apple defices through Mirai's inference engine. Same model, same output, every device.
Runs on iPhone, iPad, and Mac.
Full performance metrics on every run.
Same model, every device. Convert once.
Mirai can support your custom or non-standard layers.
We will build full conversion pipeline specifically for your model. You will have the same correctness validation and quality measurement as standard models.
Mirai can support your custom or non-standard layers.
We will build full conversion pipeline specifically for your model. You will have the same correctness validation and quality measurement as standard models.
Our optimization pipeline prepares your model for peak performance on Apple devices.
Layer fusion and
weight optimization.
Layer fusion and
weight optimization.
Layer fusion and weight optimization.
Your model is restructured and layers are fused into optimized blocks. Each block runs on the most efficient compute unit: GPU, Neural Engine, or both.
Your model is restructured and layers are fused into optimized blocks. Each block runs on the most efficient compute unit: GPU, Neural Engine, or both.
Quality degradation measurement.
Quality degradation measurement.
We measure how quantization and optimization affect your model's output quality. You see the exact tradeoff between size, speed, and accuracy.
We measure how quantization and optimization affect your model's output quality. You see the exact tradeoff between size, speed, and accuracy.
Output correctness validation.
Output correctness validation.
Output correctness validation.
Tensor-by-tensor comparison against reference outputs. Your model produces the same results after conversion. Any deviation is flagged and reported.
Tensor-by-tensor comparison against reference outputs. Your model produces the same results after conversion. Any deviation is flagged and reported.
Models already running on Mirai.
Optimize your model for 2 billion Apple devices. Perfect for:
Model companies.
Model companies.
You train and ship models. Mirai optimizes them for Apple Silicon, benchmarks on real hardware, and distributes.
You train and ship models. Mirai optimizes them for Apple Silicon, benchmarks on real hardware, and distributes.
AI researchers & labs.
AI researchers & labs.
AI researchers & labs.
Mirai converts your model and puts it in front of real users on Apple devices, not just leaderboards.
Mirai converts your model and puts it in front of real users on Apple devices, not just leaderboards.
Independent model makers.
Independent model makers.
Independent model makers.
You're fine-tuning or training from scratch. Mirai gives your model the same device reach as OpenAI and DeepSeek.
You're fine-tuning or training from scratch. Mirai gives your model the same device reach as OpenAI and DeepSeek.
Common questions:
How does model support work?
How does model support work?
What architectures are supported?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
What architectures are supported?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How does Mirai compare to other inference engines?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How does Mirai compare to other inference engines?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
What is the maximum supported model size?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
What is the maximum supported model size?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How can I run benchmarks myself?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How can I run benchmarks myself?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How can we discuss a specific use case?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
How can we discuss a specific use case?
Framer is a design tool that allows you to design websites on a freeform canvas, and then publish them as websites with a single click.
