Skip to content
llm-speed
← raw run
wrapped·Apr 28, 2026
/r/r_sx3a4y9n-m4
your run

llama-3.3-70b-Instruct-4bit

on M3 Ultra
16.8
tok/sdecode
M3 Ultra (60-core GPU) + 96GB unified
rank in tier
19/20M3 Ultra runstop 95%
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
  • phi-4-4bit on M3 Pro15.6 tok/s
  • Qwen2.5-14B-Instruct-4bit on M3 Pro14.4 tok/s
  • Qwen3-32B-4bit on M3 Pro7.2 tok/s
share
16.8 tok/s on M3 Ultra running llama-3.3-70b-Instruct-4bit — top 95% of M3 Ultra runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_sx3a4y9n-m4
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.