RTX 5070

6 records across 3 models (quants: Q4_K, UD-IQ3_XXS).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX 5070Qwen3 14B 14BQ4_Kllama.cpp54.22144.2––4096–sourced–source
RTX 5070Qwen3 14B 14BQ4_Kllama.cpp40.61315.2––16384–sourced–source
RTX 5070Qwen3 8B 8BQ4_Kllama.cpp85.83487.7––4096–sourced–source
RTX 5070Qwen3 8B 8BQ4_Kllama.cpp59.11600.8––16384–sourced–source
RTX 5070Qwen3 8B 8BQ4_Kllama.cpp43.6898.8––32768–sourced–source
RTX 5070Qwen3.8-Flash-Next UD-IQ3_XXSllama.cpp (CUDA)11.5––––2026-10-04community–source

All rows cite their source. Provenance: community 40018 sourced 1085

← all hardware