🔌
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 GBERT-Large?
Yes — here's how fast and up to what context, with real measured numbers.
GBERT-Large on the Apple M2 Ultra 64GB
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 846k | 811.9-979.1 t/s | ~10.1-12.2 (slow) |
| Q8 | yes | 851k | 1623.9-1958.2 t/s | ~20.3-24.5 (slow) |
| Q4 | yes | 853k | 2899.8-3496.8 t/s | ~36.2-43.7 (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 GBERT-Large?▶
Yes. At Q4 it generates ~2899.8-3496.8 tok/s and fits up to 853k context; beyond that it spills to SSD swap and slows to ~36.2-43.7 tok/s.
How much context fits?▶
At Q4, up to about 853k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.