🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the Apple M2 Ultra 64GB run Llama 3.2 1B?
Yes — here's how fast and up to what context, with real measured numbers.
Llama 3.2 1B on the Apple M2 Ultra 64GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 818k | 221.1-266.7 t/s | ~2.8-3.3 (slow) |
| Q8 | yes | 837k | 442.3-533.3 t/s | ~5.5-6.7 (slow) |
| Q4 | yes | 846k | 789.8-952.4 t/s | ~9.9-11.9 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.
FAQ
Can the Apple M2 Ultra 64GB run Llama 3.2 1B?▶
Yes. At Q4 it generates ~789.8-952.4 tok/s and fits up to 846k context; beyond that it spills to SSD swap and slows to ~9.9-11.9 tok/s.
How much context fits?▶
At Q4, up to about 846k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.