Skip to content
llm-speed
← raw run
wrapped·Oct 22, 2026
/r/r_--ul19zazqg
your run

Qwen3.8-27B-ctx8k

on RTX 4090
43.1
tok/sdecode
RTX 4090 (24GB)
rank in tier
61/63RTX 4090 runstop 97%
best workload
chat-short
where the rig flew
slowest workload
agent-trace42.5 tok/s
where the rig struggled
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
  • Qwen3.6-27B on RTX 309042.7 tok/s
  • Qwen3.8-27B-ctx4k on RTX 309038.6 tok/s
  • Qwen3-32B on RTX 309038.5 tok/s
share
43.1 tok/s on RTX 4090 running Qwen3.8-27B-ctx8k — top 97% of RTX 4090 runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_--ul19zazqg
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.