Run gpt-oss-20b on an RTX 5090

15 GiB needed. 32 GiB on the card. It fits.

Deploy picks the cheapest GPU that fits. 10% platform fee included, shown before you start.

Versions that fit

OpenAI's open-weight 20B reasoning model.

VersionNeeds
llama.cpp · MXFP415 GiBDeploy
vLLM · MXFP422 GiBDeploy

Other GPUs for gpt-oss-20b

Questions

Does gpt-oss-20b fit on an RTX 5090?
Yes. It needs 15 GiB of VRAM and the RTX 5090 has 32 GiB.
What does it cost to run gpt-oss-20b on an RTX 5090?
Hosts set the hourly price. You see it, with our 10% platform fee included, before anything starts. See live RTX 5090 offers.
How do I use gpt-oss-20b once it runs?
It serves an OpenAI-compatible API. The rental page gives you the URL and a key.

Go to a page or search