Skip to content
llm-speed
← raw run
wrapped·Jul 01, 2026
/r/r_b9ul-vxh9sc
your run

gpt-oss-20b

on RTX 5090
318.4
tok/sdecode
RTX 5090 (32GB) + AMD Ryzen 7 9850X3D 8-Core Processor (8c) + 30GB
rank in tier
5/183RTX 5090 runstop 3%
best workload
chat-short
where the rig flew
slowest workload
chat-long301.6 tok/s
where the rig struggled
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
  • Qwen2.5-0.5B-Instruct-4bit on M3 Pro286.5 tok/s
  • LFM2.5-24B-A2B on RTX 4090236.8 tok/s
  • Qwen3-Coder-30B-A3B-Instruct on RTX 4090232.5 tok/s
share
318.4 tok/s on RTX 5090 running gpt-oss-20b — top 3% of RTX 5090 runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_b9ul-vxh9sc
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.