filipe(@filicroval):Overnight benchmark run on GX10 for Poolside Laguna S 2.1 (Q4_K_M): - 20.4 tok/s at 1K context - 19.4 tok/s at 8K context - 11.3 tok/s at 128K context Local benchmark: - 64% coding - 60% debugging - 70% repo edits - 30% JSON - 30% review Optimized NVFP4 + vLLM/DFlash runs can produce higher throughput, especially with long generations and concurrent requests.

filipe

@filicroval

data eng | 1xAsus Ascent GX10 | benchmarking local models so you don't have to

Joined May 2023

214 Following 153.4K Followers

filipe@filicroval

2026.07.22 09:22

Overnight benchmark run on GX10 for Poolside Laguna S 2.1 (Q4_K_M): - 20.4 tok/s at 1K context - 19.4 tok/s at 8K context - 11.3 tok/s at 128K context Local benchmark: - 64% coding - 60% debugging - 70% repo edits - 30% JSON - 30% review Optimized NVFP4 + vLLM/DFlash runs can produce higher throughput, especially with long generations and concurrent requests.