RTX PRO 4000 Blackwell

7 records across 3 models (quants: EXL2 6.0bpw H6, Q4_0, Q4_K_M).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX PRO 4000 Blackwellllama 7B Q4_0llama.cpp (CUDA)136.385674.44––tg1282025-11-21community–source
RTX PRO 4000 Blackwellllama 7B Q4_0llama.cpp (CUDA)131.664992.83––tg1282025-11-21community–source
RTX PRO 4000 Blackwellllama 7B Q4_0llama.cpp (CUDA)92.544078.52––tg1282026-05-13community–source
RTX PRO 4000 Blackwellllama 7B Q4_0llama.cpp (CUDA)89.733628.68––tg1282026-05-13community–source
RTX PRO 4000 BlackwellQwen2.5-7B-Instruct EXL2 6.0bpw H6ExLlamaV247––––2026-03-16community–source
RTX PRO 4000 BlackwellQwen3.5-9B 9BQ4_K_Mllama.cpp (CUDA)90––––2026-08community–source
RTX PRO 4000 BlackwellQwen3.5-9B 9BQ4_K_Mllama.cpp (CUDA)61––––2026-08community–source

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

← all hardware