llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
NVIDIA RTX 3090
/
report
Qwen3.8 27B
on
NVIDIA RTX 3090
NVIDIA hardware
Sep 7, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
55.0 t/s gen
VRAM reported
24 GB
Summary
User reports Qwen3.8-27B at 55 tok/s on an RTX 3090. Setup is the IQ4_XS quant. The user also mentions GLM-5.3 pricing and intelligence, DeepSeek V4 Flash, and Strix Halo hardware, but does not run them.
View source ↗
Feedback