Yes — with 3.5 GB to spare
Olmo 3 7B Instruct at Q4_K_M fits your M1 · 16 GB entirely in unified memory at 8K context, at an estimated 9.3 tokens per second. Past 35K 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 · 4.1 GB
Apache 2.0
Released 20 Nov 2025
Fully open — training data and code included. Full multi-head attention, so its cache is 4× a GQA 7B at the same context.
The VRAM budget
weights 4.1 GB
Weights 4.1 GB
KV cache @ 8K 2.50 GB
Runtime overhead 0.6 GB
Free 3.5 GB of 10.7 GB
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| F16 | 13.6 GB | 16.7 GB | — | ~2.8 | Reference | 6.0 GB over |
| Q8_0 | 7.2 GB | 10.3 GB | 11K | 5.3 | −0.1% ppl | Fits |
| Q6_K | 5.6 GB | 8.7 GB | 24K | 6.8 | −0.4% ppl | Long context |
| Q5_K_M | 4.8 GB | 7.9 GB | 30K | 7.9 | −0.8% ppl | Long context |
| Q4_K_M | 4.1 GB | 7.2 GB | 35K | 9.3 | −1.9% ppl | Recommended |
| Q3_K_M | 3.3 GB | 6.4 GB | 42K | 11 | −5.4% ppl | Long context |
| Q2_K | 2.8 GB | 5.9 GB | 46K | 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. This model interleaves sliding-window layers (4096 tokens, 8 of 32 layers global), which is why its cache barely grows with context.
How to run it
$ pip install mlx-lm $ mlx_lm.generate --model mlx-community/Olmo-3-7B-Instruct-4bit \ --max-tokens 512 --prompt "Hello"
Apple's own array framework. The fastest path on Apple Silicon. More on MLX.
01Download is 4.1 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 35K context on this card.