🔌
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 1B?
Yes — here's how fast and up to what context, with real measured numbers.
Llama 3.2 1B on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1697k | 81-96.5 t/s | ~3-3.5 (slow) |
| Q8 | yes | 1716k | 162-193.1 t/s | ~5.9-7.1 (slow) |
| Q4 | yes | 1724k | 289.3-344.8 t/s | ~10.6-12.6 (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 1B?▶
Yes. At Q4 it generates ~289.3-344.8 tok/s and fits up to 1724k context; beyond that it spills to SSD swap and slows to ~10.6-12.6 tok/s.
How much context fits?▶
At Q4, up to about 1724k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.