mamba2-2.7b

6 records across 2 hardware.

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
RTX PRO 6000mamba2-2.7b 2.7BQ4_K_Mllama.cpp (CUDA)1638.9––––2026-08community–source
RTX PRO 6000mamba2-2.7b 2.7BBF16llama.cpp (CUDA)1508.0––––2026-08community–source
RTX PRO 6000mamba2-2.7b 2.7BQ4_K_Mllama.cpp (CUDA)953.0––––2026-08community–source
RTX PRO 6000mamba2-2.7b 2.7BBF16llama.cpp (CUDA)630.9––––2026-08community–source
GB10mamba2-2.7b 2.7BQ8_0llama.cpp (CUDA)246.7––––2026-08community–source
GB10mamba2-2.7b 2.7BQ8_0llama.cpp (CUDA)98.4––––2026-08community–source

← all models