RTX 4070 SUPER

7 records across 3 models (quants: Q4_0, Q4_K).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX 4070 SUPERllama 7B Q4_0llama.cpp (Vulkan)112.05198.91––tg1282026-01-16community–source
RTX 4070 SUPERllama 7B Q4_0llama.cpp (Vulkan)108.744608.2––tg1282026-01-16community–source
RTX 4070 SUPERQwen3 14B 14BQ4_Kllama.cpp45.52522.2––4096–sourced–source
RTX 4070 SUPERQwen3 14B 14BQ4_Kllama.cpp37.21578.1––16384–sourced–source
RTX 4070 SUPERQwen3 8B 8BQ4_Kllama.cpp75.44321.7––4096–sourced–source
RTX 4070 SUPERQwen3 8B 8BQ4_Kllama.cpp56.22525.8––16384–sourced–source
RTX 4070 SUPERQwen3 8B 8BQ4_Kllama.cpp42.21595.7––32768–sourced–source

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

← all hardware