5070/2x5060/4060

4 records across 1 models (quants: Q3_XXS, Q4_XS, UD-IQ4_XS).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
5070/2x5060/4060Qwen3.8-Flash-Next Q3_XXSllama.cpp (CUDA)30550––500002026-09community–source
5070/2x5060/4060Qwen3.8-Flash-Next Q4_XSllama.cpp (CUDA)20–––400002026-09community–source
5070/2x5060/4060Qwen3.8-Flash-Next UD-IQ4_XSllama.cpp (CUDA)12170––500002026-09community–source
5070/2x5060/4060Qwen3.8-Flash-Next Q3_XXSllama.cpp (CUDA)–350–––2026-09community–source

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

← all hardware