🔌
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 Belle VL?
Yes — here's how fast and up to what context, with real measured numbers.
Belle VL on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 523k | 7.1-8.5 t/s | ~0.3-0.3 (slow) |
| Q8 | yes | 609k | 14.2-17 t/s | ~0.5-0.6 (slow) |
| Q4 | yes | 646k | 25.4-30.3 t/s | ~0.9-1.1 (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 Belle VL?▶
Yes. At Q4 it generates ~25.4-30.3 tok/s and fits up to 646k context; beyond that it spills to SSD swap and slows to ~0.9-1.1 tok/s.
How much context fits?▶
At Q4, up to about 646k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.