M3 Ultra (512 GB)

16 records across 7 models (quants: 4bit, 8bit, bf16, q4.1bit).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
M3 Ultra (512 GB)altar-1 4bitMLX13.786––1282026-10-07community–source
M3 Ultra (512 GB)altar-1 4bitMLX13.0194––40962026-10-07community–source
M3 Ultra (512 GB)altar-1 4bitMLX12.6205––20482026-10-07community–source
M3 Ultra (512 GB)Devstral Small 1.1 4bitLM Studio43–––––sourced–source
M3 Ultra (512 GB)Gemma 3 27B bf16LM Studio11.19––1.72128000–sourced–source
M3 Ultra (512 GB)GLM-5 4bitMLX16.7187–5.4771024–sourced–source
M3 Ultra (512 GB)GLM-5 4bitMLX13.7180.1–22.7454096–sourced–source
M3 Ultra (512 GB)GLM-5 4bitMLX13.2154.1–53.1698192–sourced–source
M3 Ultra (512 GB)GLM-5 4bitMLX12117.4–139.54516384–sourced–source
M3 Ultra (512 GB)GLM-5 4bitMLX10.777.7–421.95532768–sourced–source
M3 Ultra (512 GB)Llama 3.3 70B 4bitLM Studio15.5150––7800–sourced–source
M3 Ultra (512 GB)Llama 3.3 70B 4bitLM Studio9.57103––39686–sourced–source
M3 Ultra (512 GB)Llama 3.3 70B 8bitLM Studio8.5150––7800–sourced–source
M3 Ultra (512 GB)Llama 3.3 70B 8bitLM Studio6.53101––39686–sourced–source
M3 Ultra (512 GB)Qwen 3 235B-A22B 4bitLM Studio27.36––1.73––sourced–source
M3 Ultra (512 GB)Qwen3.5-397B-A17B q4.1bitMLX40.2–––––sourced–source

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

← all hardware