qwen-2.5-7b

3 records across 3 hardware.

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX 4090qwen-2.5-7b Q4_K_Mllama.cpp llama-bench135–––40962026-08-17sourced–source
RTX 3090qwen-2.5-7b Q4_K_Mllama.cpp llama-bench95–––40962026-08-17sourced–source
M3 Maxqwen-2.5-7b Q4_K_Mllama.cpp llama-bench40–––40962026-08-17sourced–source

← all models