M3 Max 10+4

7 records across 1 models (quants: F16, Q4_0, Q8_0).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
M3 Max 10+4llama 7B mostly Q4_0llama.cpp (Metal)83.731030.04––tg1282024-01-11community–source
M3 Max 10+4llama 7B mostly Q8_0llama.cpp (Metal)59.871042.95––tg1282024-01-11community–source
M3 Max 10+4llama 7B mostly Q4_0llama.cpp (Metal)56.85567.59––tg1282023-12-22community–source
M3 Max 10+4llama 7B mostly F16llama.cpp (Metal)37.011168.89––tg1282024-01-11community–source
M3 Max 10+4llama 7B mostly Q8_0llama.cpp (Metal)34.3566.4––tg1282023-12-22community–source
M3 Max 10+4llama 7B mostly Q8_0llama.cpp (Metal)34.3566.4––tg1282024-01-11communitycontradictionsource
M3 Max 10+4llama 7B mostly F16llama.cpp (Metal)19.54589.41––tg1282023-12-22community–source

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

← all hardware