Qwen3-32B

on RTX 4090

43.5

tok/sdecode

RTX 4090 (24GB) + AMD EPYC 7642 48-Core Processor (48c) + 252GB

rank in tier

29/31RTX 4090 runstop 94%

best workload

chat-short

where the rig flew

slowest workload

chat-long40.7 tok/s

where the rig struggled

backend

llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.

faster than

Qwen3.8-27B on RTX 309037.6 tok/s
Qwen2.5-32B-Instruct-4bit on M3 Ultra34.6 tok/s
Qwen2.5-Coder-32B-Instruct-4bit on M3 Ultra34.5 tok/s

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.