V100

26 records across 5 models (quants: Q4_0, Q4_K - Medium).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
V100gemma3 12B Q4_K - Mediumllama.cpp (CUDA)68.941771.04––tg1282025-09-22community–source
V100gemma3 12B Q4_K - Mediumllama.cpp (CUDA)67.341857.73––tg1282025-09-22community–source
V100llama 7B Q4_0llama.cpp (CUDA)152.363134.81––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)152.333537.71––tg1282025-12-17community–source
V100llama 7B Q4_0llama.cpp (CUDA)151.883133.71––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)150.683075.88––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)145.453434.83––tg1282025-12-17community–source
V100llama 7B Q4_0llama.cpp (CUDA)140.233261.46––tg1282025-12-17community–source
V100llama 7B Q4_0llama.cpp (CUDA)140.03235.72––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)139.893235.01––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)139.113149.71––tg1282025-11-19community–source
V100llama 7B Q4_0llama.cpp (CUDA)134.762973.78––tg1282025-09-22community–source
V100llama 7B Q4_0llama.cpp (CUDA)133.063160.43––tg1282025-12-17community–source
V100llama 7B Q4_0llama.cpp (CUDA)129.083042.64––tg1282025-09-22community–source
V100llama 7B Q4_0llama.cpp (CUDA)109.49380.4––tg1282026-05-26community–source
V100llama 7B Q4_0llama.cpp (CUDA)109.461156.76––tg1282026-05-26community–source
V100llama 7B Q4_0llama.cpp (CUDA)99.95977.33––tg1282026-05-26community–source
V100llama 7B Q4_0llama.cpp (CUDA)99.84354.07––tg1282026-05-26community–source
V100llama 7B Q4_0llama.cpp (CUDA)62.06788.45––tg1282025-11-26community–source
V100llama 7B Q4_0llama.cpp (CUDA)58.33767.12––tg1282025-11-26community–source
V100qwen2 70B Q4_0llama.cpp (CUDA)17.72324.02––tg1282025-09-22community–source
V100qwen2 70B Q4_0llama.cpp (CUDA)17.22335.74––tg1282025-09-22community–source
V100Qwen3 14B Q4_K - Mediumllama.cpp (CUDA)67.031474.29––tg1282025-09-22community–source
V100Qwen3 14B Q4_K - Mediumllama.cpp (CUDA)64.871536.27––tg1282025-09-22community–source
V100Qwen3 32B Q4_K - Mediumllama.cpp (CUDA)31.8613.13––tg1282025-09-22community–source
V100Qwen3 32B Q4_K - Mediumllama.cpp (CUDA)30.74642.41––tg1282025-09-22community–source

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

← all hardware