🔌
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 Mamba2-Hybrid?
Yes — here's how fast and up to what context, with real measured numbers.
Mamba2-Hybrid on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 735k | 11.5-13.7 t/s | ~0.4-0.5 (slow) |
| Q8 | yes | 801k | 23-27.4 t/s | ~0.8-1 (slow) |
| Q4 | yes | 830k | 41.1-49 t/s | ~1.5-1.8 (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 Mamba2-Hybrid?▶
Yes. At Q4 it generates ~41.1-49 tok/s and fits up to 830k context; beyond that it spills to SSD swap and slows to ~1.5-1.8 tok/s.
How much context fits?▶
At Q4, up to about 830k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.