New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Apple M4 Max 48GB
Runs downloadable LLMs up to ~69.5B at Q4, ~82.9-99.9 tok/s on an 8B model.
What you can run
At 8k context, FP16 KV. Full-speed tok/s holds up to "max ctx"; beyond it the context spills to SSD swap and drops to the degraded figure.
| Model size | FP16 | Q8 | Q4 |
|---|---|---|---|
| 3B |
61.9-74.6 t/s ≤545k ctx · then ~1.1-1.4 (slow) |
123.8-149.2 t/s ≤591k ctx · then ~2.3-2.7 (slow) |
221-266.5 t/s ≤611k ctx · then ~4-4.9 (slow) |
| 7-8B |
23.2-28 t/s ≤196k ctx · then ~0.4-0.5 (slow) |
46.4-56 t/s ≤257k ctx · then ~0.9-1 (slow) |
82.9-99.9 t/s ≤284k ctx · then ~1.5-1.8 (slow) |
| 13B |
14.3-17.2 t/s ≤96k ctx · then ~0.3-0.3 (slow) |
28.6-34.4 t/s ≤175k ctx · then ~0.5-0.6 (slow) |
51-61.5 t/s ≤210k ctx · then ~0.9-1.1 (slow) |
| 32B | won't run |
11.6-14 t/s ≤37k ctx · then ~0.2-0.3 (slow) |
20.7-25 t/s ≤91k ctx · then ~0.4-0.5 (slow) |
| 70B | won't run | won't run |
9.5-11.4 t/s ≤8k ctx · then ~0.2-0.2 (slow) |
| 120B | won't run | won't run | won't run |
| 400B | won't run | won't run | won't run |
Estimate: tok/s ≈ bandwidth ÷ (params × bytes/param) × efficiency (Apple). Cyan = fits ~43.2 GB usable. Beyond "ctx", KV spills to SSD swap (rough degraded speed shown).
Run popular models on the Apple M4 Max 48GB
Videos about the Apple M4 Max 48GB
Qwen3 Coder M4 Max vs RTX 5090
mlx vs ollama on m4 max macbook pro
More hardware like the Apple M4 Max 48GB
About the Apple M4 Max 48GB
Apple M4 Max 48GB is a unified-memory system from Apple on Apple M4, 2024. For local genAI the numbers that matter are its 48 GB (how big a model + context fits) and 546 GB/s bandwidth (generation speed).
Frequently asked questions
What size LLM can the Apple M4 Max 48GB run?▶
With ~43.2 GB usable it runs up to ~69.5B at Q4 (~19.5B at FP16) at 8k context. Bigger spills into SSD swap and slows sharply.
How many tokens/sec does the Apple M4 Max 48GB generate?▶
Generation is bandwidth-bound. At 546 GB/s an 8B model at Q4 runs ~82.9-99.9 tok/s (estimate).
Why does long context slow it down?▶
The KV cache for the context must also fit fast memory; once weights+KV exceed 43.2 GB it spills to SSD swap and throughput collapses. Each row shows the max usable context.