🔌
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 DEQ-Transformer (Post-LN) + Jacobian Regularisation?
Yes — here's how fast and up to what context, with real measured numbers.
DEQ-Transformer (Post-LN) + Jacobian Regularisation on the Apple M2 Ultra 64GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 853k | 2775.5-3346.9 t/s | ~34.7-41.8 (slow) |
| Q8 | yes | 855k | 5551-6693.9 t/s | ~69.4-83.7 (slow) |
| Q4 | yes | 855k | 9912.5-11953.4 t/s | ~123.9-149.4 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.
→ All models on the Apple M2 Ultra 64GB
→ DEQ-Transformer (Post-LN) + Jacobian Regularisation specs
→ Other hardware
FAQ
Can the Apple M2 Ultra 64GB run DEQ-Transformer (Post-LN) + Jacobian Regularisation?▶
Yes. At Q4 it generates ~9912.5-11953.4 tok/s and fits up to 855k context; beyond that it spills to SSD swap and slows to ~123.9-149.4 tok/s.
How much context fits?▶
At Q4, up to about 855k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.