M3 Ultra 24+8

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

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
M3 Ultra 24+8llama 7B mostly Q4_0llama.cpp (Metal)92.141471.24––tg1282025-03-23community–source
M3 Ultra 24+8llama 7B mostly Q8_0llama.cpp (Metal)63.931487.51––tg1282025-03-23community–source
M3 Ultra 24+8llama 7B mostly F16llama.cpp (Metal)39.781538.34––tg1282025-03-23community–source

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

← all hardware