🔌
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 MiniGPT4 (Vicuna finetune)?
Yes — here's how fast and up to what context, with real measured numbers.
MiniGPT4 (Vicuna finetune) on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 535k | 7.7-9.1 t/s | ~0.3-0.3 (slow) |
| Q8 | yes | 615k | 15.3-18.3 t/s | ~0.6-0.7 (slow) |
| Q4 | yes | 650k | 27.4-32.6 t/s | ~1-1.2 (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 MiniGPT4 (Vicuna finetune)?▶
Yes. At Q4 it generates ~27.4-32.6 tok/s and fits up to 650k context; beyond that it spills to SSD swap and slows to ~1-1.2 tok/s.
How much context fits?▶
At Q4, up to about 650k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.