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

gemma-3-4b-it

by google

4.3B params · image-text-to-text · 1.4k likes · 1.6M downloads

gemma-3-4b-it is a 4.3B parameter model. At Q4 quantization it requires 2GB of VRAM. It runs comfortably on GeForce RTX 4090 (305 tok/s), GeForce RTX 5090 (457 tok/s), M4 Max 128GB (111 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
Featherless
deepinfra39 tok/s

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB305 tok/scomfortable
GeForce RTX 509032GB457 tok/scomfortable
M4 Max 128GB128GB111 tok/scomfortable
M4 Pro 48GB48GB55 tok/scomfortable
M4 Pro 24GB24GB55 tok/scomfortable
A100 PCIe 80 GB80GB559 tok/scomfortable
H100 SXM5 80 GB80GB1125 tok/scomfortable
GeForce RTX 309024GB269 tok/scomfortable
Radeon RX 7900 XTX24GB223 tok/scomfortable
GeForce RTX 408016GB216 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0