How we will benchmark local LLM performance, and theoretical bandwidth-based speed ceilings in the meantime. We only publish measured results we can document or cite from a reproducible public run.
2 articles
Local LLM Benchmarks
Local LLM benchmark hub: how we will test tokens per second across GPUs and mini PCs, and theoretical bandwidth…
Read →DGX Spark Benchmarks
DGX Spark benchmark hub: methodology and results for local LLM performance, published only from verified…
Read →