Skip to content
llm-speed
wrapped·Nov 06, 2026
/r/r_-jalsgvd-br
your run

Qwen3.8-27B

25.2
tok/sdecode
M3 Ultra (60-core GPU) + 96GB unified
rank in tier
26/30M3 Ultra runstop 87%
best workload
chat-short
where the rig flew
slowest workload
—
single-workload run
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
  • llama3.2 on AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD23.9 tok/s
  • Muse-Glimmer-30B-ctx8k on RTX A400023.3 tok/s
  • gemma-2-9b-it-4bit on M3 Pro22.0 tok/s
share
25.2 tok/s on M3 Ultra running Qwen3.8-27B — top 87% of M3 Ultra runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_-jalsgvd-br

This is a shareable view of one signed llm-speed suite-suite-v1 run. Numbers are not edited. The raw record has every workload, the public key, and the fingerprint hash.