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

roberta-base

by FacebookAI

125M params · fill-mask · 567 likes · 10.7M downloads

roberta-base is a 125M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (10529 tok/s), GeForce RTX 5090 (15790 tok/s), M4 Max 128GB (3853 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
HF Inference

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB10529 tok/scomfortable
GeForce RTX 509032GB15790 tok/scomfortable
M4 Max 128GB128GB3853 tok/scomfortable
M4 Pro 48GB48GB1926 tok/scomfortable
M4 Pro 24GB24GB1926 tok/scomfortable
A100 PCIe 80 GB80GB19291 tok/scomfortable
H100 SXM5 80 GB80GB38800 tok/scomfortable
GeForce RTX 309024GB9309 tok/scomfortable
Radeon RX 7900 XTX24GB7698 tok/scomfortable
GeForce RTX 408016GB7472 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0