llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
M3 Ultra
/
report
DeepSeek V4 Flash 284B (13B active)
on M3 Ultra · MLX
Tone: positive
Sep 7, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
43.0 t/s gen
Quant
MXFP4
System RAM
192 GB
Summary
User reports 34 t/s at the start of a run, rising to 43 t/s by the end. The cached tokens were the default chat prompt and the query was 13k.
View source ↗
Feedback