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

resnet-50

by microsoft

26M params · image-classification · 495 likes · 344.8k downloads

resnet-50 is a 26M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (51268 tok/s), GeForce RTX 5090 (76883 tok/s), M4 Max 128GB (18761 tok/s).

Inference providers

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

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB51268 tok/scomfortable
GeForce RTX 509032GB76883 tok/scomfortable
M4 Max 128GB128GB18761 tok/scomfortable
M4 Pro 48GB48GB9380 tok/scomfortable
M4 Pro 24GB24GB9380 tok/scomfortable
A100 PCIe 80 GB80GB93931 tok/scomfortable
H100 SXM5 80 GB80GB188925 tok/scomfortable
GeForce RTX 309024GB45329 tok/scomfortable
Radeon RX 7900 XTX24GB37485 tok/scomfortable
GeForce RTX 408016GB36385 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0