boltbeam.fyi

Qwen3-14B speed on every GPU

11 measured runs across 6 GPUs, fastest first.

GPUQuantEnginetok/sRooflineShare
RTX 5090AWQvLLM153.0200.976%
RTX 5090Q4_K_Mllama.cpp149.5197.876%
H100 80GB HBM3AWQvLLM147.6380.339%
H100 80GB HBM3Q4_K_Mllama.cpp144.4374.339%
H200 NVLQ4_K_Mllama.cpp135.0521.826%
RTX 4090AWQvLLM97.7113.086%
RTX 4090Q4_K_Mllama.cpp94.4111.285%
RTX 3090AWQvLLM82.4105.478%
L40SAWQvLLM77.789.487%
RTX 3090Q4_K_Mllama.cpp76.4103.874%
L40SQ4_K_Mllama.cpp74.188.184%

Download