🔌
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 Mamba-2.8B?
Yes — here's how fast and up to what context, with real measured numbers.
Mamba-2.8B on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1649k | 35.6-42.4 t/s | ~1.3-1.6 (slow) |
| Q8 | yes | 1692k | 71.2-84.8 t/s | ~2.6-3.1 (slow) |
| Q4 | yes | 1711k | 127.1-151.5 t/s | ~4.7-5.5 (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 Mamba-2.8B?▶
Yes. At Q4 it generates ~127.1-151.5 tok/s and fits up to 1711k context; beyond that it spills to SSD swap and slows to ~4.7-5.5 tok/s.
How much context fits?▶
At Q4, up to about 1711k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.