vram.run Models Hardware Providers Cloud State of Inference
API provider data is live · Hardware & cloud pricing curated 2026-02-23

gpt-oss-20b

by openai

21.5B params · text-generation · 4.4k likes · 7.4M downloads

gpt-oss-20b is a 21.5B parameter model. At Q4 quantization it requires 11GB of VRAM. It runs comfortably on GeForce RTX 4090 (61 tok/s), GeForce RTX 5090 (91 tok/s), A100 PCIe 80 GB (111 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
Groq558 tok/s
Novita97 tok/s
Nscale121 tok/s
Hyperbolic95 tok/s
Together AI76 tok/s
Fireworks143 tok/s
OVHcloud85 tok/s

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB61 tok/scomfortable
GeForce RTX 509032GB91 tok/scomfortable
M4 Max 128GB128GB22 tok/stight
M4 Pro 48GB48GB11 tok/stight
M4 Pro 24GB24GB11 tok/stight
A100 PCIe 80 GB80GB111 tok/scomfortable
H100 SXM5 80 GB80GB224 tok/scomfortable
GeForce RTX 309024GB53 tok/scomfortable
Radeon RX 7900 XTX24GB44 tok/scomfortable
GeForce RTX 408016GB43 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0