🔌
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 M4 Max 48GB run Adaptive Inputs + LayerDrop?
Yes — here's how fast and up to what context, with real measured numbers.
Adaptive Inputs + LayerDrop on the Apple M4 Max 48GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 623k | 438.9-529.2 t/s | ~8-9.7 (slow) |
| Q8 | yes | 630k | 877.7-1058.4 t/s | ~16.1-19.4 (slow) |
| Q4 | yes | 633k | 1567.4-1890.1 t/s | ~28.7-34.6 (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 M4 Max 48GB run Adaptive Inputs + LayerDrop?▶
Yes. At Q4 it generates ~1567.4-1890.1 tok/s and fits up to 633k context; beyond that it spills to SSD swap and slows to ~28.7-34.6 tok/s.
How much context fits?▶
At Q4, up to about 633k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.