🔌
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 M3 Max 64GB run Selfish-RNN (SNT-ASGD) Stacked LSTMs?
Yes — here's how fast and up to what context, with real measured numbers.
Selfish-RNN (SNT-ASGD) Stacked LSTMs on the Apple M3 Max 64GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 855k | 5396.8-6507.9 t/s | ~134.9-162.7 (slow) |
| Q8 | yes | 856k | 10793.7-13015.9 t/s | ~269.8-325.4 (slow) |
| Q4 | yes | 856k | 19274.4-23242.6 t/s | ~481.9-581.1 (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 M3 Max 64GB run Selfish-RNN (SNT-ASGD) Stacked LSTMs?▶
Yes. At Q4 it generates ~19274.4-23242.6 tok/s and fits up to 856k context; beyond that it spills to SSD swap and slows to ~481.9-581.1 tok/s.
How much context fits?▶
At Q4, up to about 856k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.