RTX 4090M

8 records across 1 models (quants: Q8_0).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)52.7–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)52.4–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)52.0–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)49.0–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)48.2–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)44.9–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)40.3–––327682026-04-01community–source
RTX 4090MQwen 3.5 9B Q8_0llama.cpp (CUDA)34.5–––327682026-04-01community–source

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

← all hardware