llamaperf
Calculator
GPUs
Engines
Guides
Blog
Community
MCP
Advertise
Submit
Sign in
Submit
llamaperf
/
NVIDIA RTX 5090
/
report
Qwen3.8 125B (6B active) Flash-Next
on
NVIDIA RTX 5090
NVIDIA hardware
Sep 7, 2026
Share
X
Reddit
LinkedIn
Bluesky
Copy
link
Throughput
43.4 t/s gen
VRAM reported
32 GB
Summary
User reports running unsloth/Qwen3.8-Flash-Next-GGUF on an RTX 5090. The user asks how it runs on the 5090 and hopes for more benchmarks.
View source ↗
Feedback