MLX
Apple's own array framework. The native path on Apple Silicon, with its own model format.
From the file· capabilities, not benchmarks
What it is
A framework built for unified memory, with mlx-lm providing model loading and generation. Because it is designed around Apple's architecture rather than ported to it, it often gets more out of the same chip than a general-purpose engine.
Capabilities
| Model formats | MLX (safetensors-based) |
| KV cache quantization | yes |
| CPU offload | not supported |
| MoE expert offload | not supported |
| Multi-GPU | Not applicable — one chip, one unified memory pool. |
| Concurrency | Single-user focused. |
| Platforms | macOS |
Best for
Apple Silicon, particularly at larger model sizes where the unified memory pool is the whole reason you bought the machine.
Watch out for
Its models are a separate format, so a GGUF you already downloaded will not work. Model availability is narrower than GGUF, though the popular families are all converted.