Skip to content
llm-speed
← raw run
wrapped·Apr 28, 2026
/r/r_anmmc80-aoq
your run

Qwen3-32B-4bit

on M3 Ultra
34.4
tok/sdecode
M3 Ultra (60-core GPU) + 96GB unified
rank in tier
24/26M3 Ultra runstop 92%
best workload
chat-short
where the rig flew
slowest workload
—
single-workload run
backend
mlxMLX is the fastest backend on Apple Silicon for dense models.
faster than
  • Qwen2.5-7B-Instruct-4bit on M3 Pro30.5 tok/s
  • Llama-3.1-8B-Instruct-4bit on M3 Pro29.2 tok/s
  • llama3.2 on AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD23.9 tok/s
share
34.4 tok/s on M3 Ultra running Qwen3-32B-4bit — top 92% of M3 Ultra runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_anmmc80-aoq
X / TwitterBlueskyMastodonRedditraw run

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.