Skip to content
llm-speed
← raw run
wrapped·Oct 29, 2026
/r/r_f0etksoa6-y
your run

Qwen3.8-27B-filled16k

on RTX 4090
41.2
tok/sdecode
RTX 4090 (24GB)
rank in tier
68/70RTX 4090 runstop 97%
best workload
long-context-decay
where the rig flew
slowest workload
—
single-workload run
backend
llama.cppllama.cpp is the universal backend — broad model support, modest speed ceiling.
faster than
  • Qwen3.8-27B-ctx8k on RTX 309038.8 tok/s
  • Qwen3-32B on RTX 309038.5 tok/s
  • Qwen2.5-32B-Instruct-4bit on M3 Ultra34.6 tok/s
share
41.2 tok/s on RTX 4090 running Qwen3.8-27B-filled16k — top 97% of RTX 4090 runs. Signed benchmark via llm-speed.
https://llm-speed.com/wrapped/r_f0etksoa6-y
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.