🔌
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 Pangu-Weather?
Yes — here's how fast and up to what context, with real measured numbers.
Pangu-Weather on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1727k | 389.2-463.9 t/s | ~14.3-17 (slow) |
| Q8 | yes | 1731k | 778.5-927.8 t/s | ~28.5-34 (slow) |
| Q4 | yes | 1733k | 1390.1-1656.7 t/s | ~50.9-60.7 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.
FAQ
Can the NVIDIA DGX Spark run Pangu-Weather?▶
Yes. At Q4 it generates ~1390.1-1656.7 tok/s and fits up to 1733k context; beyond that it spills to SSD swap and slows to ~50.9-60.7 tok/s.
How much context fits?▶
At Q4, up to about 1733k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.