🔌
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 Llama 3.2 3B?
Yes — here's how fast and up to what context, with real measured numbers.
Llama 3.2 3B on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1637k | 31-37 t/s | ~1.1-1.4 (slow) |
| Q8 | yes | 1686k | 62.1-74 t/s | ~2.3-2.7 (slow) |
| Q4 | yes | 1707k | 110.9-132.1 t/s | ~4.1-4.8 (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 Llama 3.2 3B?▶
Yes. At Q4 it generates ~110.9-132.1 tok/s and fits up to 1707k context; beyond that it spills to SSD swap and slows to ~4.1-4.8 tok/s.
How much context fits?▶
At Q4, up to about 1707k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.