RTX 2070 Super

7 records across 2 models (quants: Q4_0, UD-Q4_K_XL).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)46.35––––2026-09community–source
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)38.33––––2026-09community–source
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)38.10––––2026-09community–source
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)37.02––––2026-09community–source
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)36.73––––2026-09community–source
RTX 2070 Supergemma-4-26B-A4B-it 26BUD-Q4_K_XLllama.cpp (CUDA)35.36––––2026-09community–source
RTX 2070 Superllama 7B Q4_0llama.cpp (Vulkan)64.641199.13––tg1282025-02-07community–source

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

← all hardware