Skip to content
llm-speed
← raw run
wrapped·Oct 22, 2026
/r/r_acy-a8qic5j
your run

Flint-40B-A2.5B-UD-cpumoe12-ctx8k

on RTX A4000
15.7
tok/sdecode
RTX A4000 (16GB) + AMD EPYC 7453 28-Core Processor (56c) + 504GB
rank in tier
18/18RTX A4000 runstop 100%
best workload
chat-short
where the rig flew
slowest workload
chat-long10.0 tok/s
where the rig struggled
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
  • phi-4-4bit on M3 Pro15.6 tok/s
  • Qwen2.5-14B-Instruct-4bit on M3 Pro14.4 tok/s
  • Llama-3.3-70B-Instruct on A4013.0 tok/s
share
15.7 tok/s on RTX A4000 running Flint-40B-A2.5B-UD-cpumoe12-ctx8k — top 100% of RTX A4000 runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_acy-a8qic5j
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.