llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
M1 Ultra 128GB
/
report
DeepSeek V4
on
M1 Ultra 128GB
· LM Studio
Tone: positive
Apple hardware
Sep 7, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
16.0 t/s gen
Quant
IQ3_XXS
System RAM
128 GB
Summary
User reports a patched engine improved speed from 5-6 to 15-16 tok/s. Setup uses the Unsloth UD-IQ3_XXS quant. The user notes a wired limit at 120GB.
View source ↗
Feedback