Yes — with 5.4 GB to spare
Qwen2.5-Coder 7B at Q4_K_M fits your M1 · 16 GB entirely in unified memory at 8K context, at an estimated 8.9 tokens per second. Past 106K the KV cache pushes it over — quantise the cache to q8_0, or step down a quantisation, to go longer.
Fully in unified memory
8K context
Q4_K_M · 4.3 GB
Apache 2.0
Released Nov 2024
The standard local autocomplete model — small enough to keep resident all day, and still the best FIM model under 8B.
The VRAM budget
weights 4.3 GB
Weights 4.3 GB
KV cache @ 8K 0.44 GB
Runtime overhead 0.6 GB
Free 5.4 GB of 10.7 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 14.2 GB | 15.2 GB | — | ~2.7 | Reference | 4.5 GB over |
| Q8_0 | 7.5 GB | 8.6 GB | 46K | 5.0 | −0.1% ppl | Long context |
| Q6_K | 5.8 GB | 6.9 GB | 78K | 6.5 | −0.4% ppl | Long context |
| Q5_K_M | 5.0 GB | 6.1 GB | 92K | 7.6 | −0.8% ppl | Long context |
| Q4_K_M | 4.3 GB | 5.3 GB | 106K | 8.9 | −1.9% ppl | Recommended |
| Q3_K_M | 3.5 GB | 4.5 GB | 121K | 11 | −5.4% ppl | Long context |
| Q2_K | 3.0 GB | 4.0 GB | 128K | 13 | −15% ppl | Long context |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.
How to run it
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/Qwen2.5-Coder-7B-Instruct-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 4.3 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02macOS caps what the GPU may wire down at about 10.7 GB of 16 GB. Raising it with iogpu.wired_limit_mb is possible, and risky.
03There is room to go to 106K context on this card.