🔌
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 TRIMELMlong (150M)?
Yes — here's how fast and up to what context, with real measured numbers.
TRIMELMlong (150M) on the NVIDIA DGX Spark
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1730k | 664.3-791.7 t/s | ~24.3-29 (slow) |
| Q8 | yes | 1733k | 1328.6-1583.4 t/s | ~48.7-58 (slow) |
| Q4 | yes | 1734k | 2372.5-2827.5 t/s | ~86.9-103.6 (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 TRIMELMlong (150M)?▶
Yes. At Q4 it generates ~2372.5-2827.5 tok/s and fits up to 1734k context; beyond that it spills to SSD swap and slows to ~86.9-103.6 tok/s.
How much context fits?▶
At Q4, up to about 1734k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.