All IntegrationsLLM Providers
MLX + Grip OS
MLX is Apple's machine learning framework optimized for Apple Silicon. Grip OS uses MLX for ultra-low-latency local inference that powers features like Grip Mail's self-learning email classification. With MLX, models run natively on the Metal GPU with unified memory access, delivering sub-10ms response times that make AI feel instantaneous. Use MLX for routine classification, embedding generation, and fine-tuning workflows that stay entirely on your device.
How to Connect MLX
1
MLX is included with Grip OS — no separate installation needed.
2
Open Settings > Model Providers to see available MLX models.
3
Download additional MLX-format models from Hugging Face via the model manager.
4
MLX models appear alongside cloud and Ollama models in the model picker.
Available Tools
model_selectlocal_inferencefine_tuneembedding_generatemodel_list
via gripos-mcp
Popular Workflows with MLX
Automation Recipes
Frequently Asked Questions
How is MLX different from Ollama?
MLX runs directly on Apple's Metal GPU framework with unified memory, offering lower latency than Ollama on Apple Silicon. Ollama supports more model formats and architectures.
Can I fine-tune models with MLX?
Yes. Grip OS supports LoRA fine-tuning through MLX for personalizing models to your workflow. This powers Grip Mail's self-learning email classification.
What hardware do I need for MLX?
Any Apple Silicon Mac (M1 or later). 16GB RAM runs small models well. 32GB+ recommended for larger models and fine-tuning.