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

Qwen2.5-7B-Instruct-4bit

on M3 Pro
30.5
tok/sdecode
M3 Pro (18-core GPU) + 36GB unified
rank in tier
12/27M3 Pro runstop 44%
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
  • claude-sonnet-4.5 on RTX 509021.6 tok/s
  • llama-3.3-70b-Instruct-4bit on M3 Ultra16.8 tok/s
  • qwen2.5-72b-Instruct-4bit on M3 Ultra16.3 tok/s
share
30.5 tok/s on M3 Pro running Qwen2.5-7B-Instruct-4bit — top 44% of M3 Pro runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_llzv_g-ymaf
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.