Yes — with 7.4 GB to spare
Llama 3.2 3B Instruct at Q4_K_M fits your M1 · 16 GB entirely in unified memory at 8K context, at an estimated 21 tokens per second. Past 75K 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 · 1.8 GB
Llama 3.2 Community
Released Sep 2024
The 2024 "it just runs" model for 8 GB laptops. Qwen3.5 4B does the same job better now.
The VRAM budget
weights 1.8 GB
Weights 1.8 GB
KV cache @ 8K 0.88 GB
Runtime overhead 0.6 GB
Free 7.4 GB of 10.7 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 6.0 GB | 7.5 GB | 37K | 6.4 | Reference | Long context |
| Q8_0 | 3.2 GB | 4.7 GB | 63K | 12 | −0.1% ppl | Long context |
| Q6_K | 2.5 GB | 3.9 GB | 69K | 16 | −0.4% ppl | Long context |
| Q5_K_M | 2.1 GB | 3.6 GB | 72K | 18 | −0.8% ppl | Long context |
| Q4_K_M | 1.8 GB | 3.3 GB | 75K | 21 | −1.9% ppl | Recommended |
| Q3_K_M | 1.5 GB | 2.9 GB | 78K | 26 | −5.4% ppl | Long context |
| Q2_K | 1.3 GB | 2.7 GB | 80K | 30 | −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/Llama-3.2-3B-Instruct-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 1.8 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 75K context on this card.