MLX on the M4 16GB
On the M4 16GB, MLX runs Mellum2.1 12B · 2.5B active at 4-bit at a median 64.0 tokens per second for generation, from 1 plain run (1K context).
Speed by model on the M4 16GB
Plain runs only: one device, one request, no speculative decoding, the whole model in memory. Each row is one model size at one bit class, with the other engines' plain runs of the same model on this card beside it. Context length and build still differ between runs, and each side shows its own.
| Model | MLX | Other engines |
|---|---|---|
| Mellum2.1 12B · 2.5B active 4-bit | 64.0 t/sone report, 1K context | no other engine reported |
| Qwen3 8B 4-bit | 10.0 t/sone report, context not stated | no other engine reported |
4 MLX reports on the M4 16GB
Newest first. A run on several cards, with speculative decoding, with part of the model in system RAM or with batched requests says so, since each describes a different setup.
- Qwen3 · 4-bit12.7 t/s
- Qwen3 · 4-bit23.2 t/s
- Mellum2.1 12B · 2.5B active · NVFP464.0 t/s
- Qwen3 8B · 4bit10.0 t/s
Frequently asked
How fast is MLX on the M4 16GB?
On the M4 16GB, MLX runs Mellum2.1 12B · 2.5B active at 4-bit at a median 64.0 tokens per second for generation, from 1 plain run (1K context).