🔌
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 Max 96GB run MT-DNN?
Yes — here's how fast and up to what context, with real measured numbers.
MT-DNN on the Apple M2 Max 96GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1285k | 412.1-497 t/s | ~10.3-12.4 (slow) |
| Q8 | yes | 1290k | 824.2-993.9 t/s | ~20.6-24.8 (slow) |
| Q4 | yes | 1293k | 1471.9-1774.9 t/s | ~36.8-44.4 (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 Max 96GB run MT-DNN?▶
Yes. At Q4 it generates ~1471.9-1774.9 tok/s and fits up to 1293k context; beyond that it spills to SSD swap and slows to ~36.8-44.4 tok/s.
How much context fits?▶
At Q4, up to about 1293k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.