🔌
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 NVIDIA DGX Spark run Ankh_large?
Yes — here's how fast and up to what context, with real measured numbers.
Ankh_large on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1677k | 52.4-62.5 t/s | ~1.9-2.3 (slow) |
| Q8 | yes | 1706k | 104.9-125 t/s | ~3.8-4.6 (slow) |
| Q4 | yes | 1719k | 187.3-223.2 t/s | ~6.9-8.2 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.
FAQ
Can the NVIDIA DGX Spark run Ankh_large?▶
Yes. At Q4 it generates ~187.3-223.2 tok/s and fits up to 1719k context; beyond that it spills to SSD swap and slows to ~6.9-8.2 tok/s.
How much context fits?▶
At Q4, up to about 1719k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.