🔌
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 Codestral Mamba?
Yes — here's how fast and up to what context, with real measured numbers.
Codestral Mamba on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 756k | 13.7-16.3 t/s | ~0.5-0.6 (slow) |
| Q8 | yes | 812k | 27.4-32.6 t/s | ~1-1.2 (slow) |
| Q4 | yes | 836k | 48.8-58.2 t/s | ~1.8-2.1 (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 Codestral Mamba?▶
Yes. At Q4 it generates ~48.8-58.2 tok/s and fits up to 836k context; beyond that it spills to SSD swap and slows to ~1.8-2.1 tok/s.
How much context fits?▶
At Q4, up to about 836k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.