Free Download — Apple Silicon

MINT-UI.

Quantize any LLM. Run it on your Mac. Keep everything local.

The individual edition of our MINT framework — a complete desktop app that takes any HuggingFace model, compresses it with mixed-precision quantization, and serves it locally with an OpenAI-compatible API. All on Apple Silicon.

From HuggingFace to local inference in six steps.

MINT-UI wraps our entire MINT quantization pipeline into a guided wizard. Pick a model, set your memory budget, and the app handles sensitivity analysis, optimal bit-width allocation, conversion, and serving.

1

Choose a Model

Pick any LLM from HuggingFace or load one from your local cache. Supports dense and Mixture-of-Experts architectures.

2

Set Your Memory Budget

Tell MINT how much memory you have. It analyses every tensor and finds the optimal mixed-precision allocation for your hardware.

3

Quantize & Serve

One click to compress. One click to serve. Get an OpenAI-compatible API and built-in chat interface — all running locally on your Mac.

What MINT-UI can do.

MINT Wizard

A 6-step guided workflow that takes you from model selection through sensitivity analysis, optimisation, format conversion, and serving. No command line. No configuration files.

Interactive Quality Charts

Visualise the quality-vs-size tradeoff in real time. MINT automatically detects the optimal knee point — or drag the slider to choose your own balance.

MLX & GGUF Output

Convert to MLX format for native Apple Silicon performance, or GGUF for llama.cpp compatibility. Your quantized model, your choice of runtime.

Local Model Server

Serve any quantized model with an OpenAI-compatible API endpoint. Built-in chat interface included. Connect your own tools, or just talk to the model.

Memory Intelligence

Automatic KV cache estimation, resource warnings, and memory-aware recommendations. MINT-UI knows your hardware and optimises accordingly.

Session Continuity

Resume any pipeline step without re-running earlier analysis. Close the app, come back tomorrow — your work is exactly where you left it.

Individual edition vs Enterprise.

MINT-UI is the free individual edition — perfect for developers, researchers, and teams who want to evaluate MINT on their own hardware. For production deployments, our enterprise platform adds the controls organisations need.

Free — Open Source

MINT-UI Individual

  • Full MINT quantization pipeline
  • Interactive wizard & quality charts
  • MLX & GGUF export
  • Local model serving & chat
  • Single-user, single-machine
Enterprise

MINT Enterprise Platform

  • Everything in Individual, plus:
  • CI/CD pipeline integration — automated quantization on model release
  • Multi-user access controls, audit logging, and role-based permissions
  • Fleet deployment across Apple Silicon clusters and on-prem infrastructure
  • Model registry, versioning, and automated benchmarking
  • Air-gapped and sovereign deployment support
  • Dedicated engineering support and SLA
Talk to Us About Enterprise

Try it yourself.

Pick a model. See what MINT quantization can do on your Mac. No sign-up. No cloud. No strings.

When you're ready to scale it across your organisation, we're here.

Enterprise Enquiry