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

madlad400-3b-mt

by google

2.9B params · translation · 208 likes · 197.4k downloads

madlad400-3b-mt is a 2.9B parameter model. At Q4 quantization it requires 1GB of VRAM. It runs comfortably on GeForce RTX 4090 (446 tok/s), GeForce RTX 5090 (669 tok/s), M4 Max 128GB (163 tok/s).

Inference providers

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

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB446 tok/scomfortable
GeForce RTX 509032GB669 tok/scomfortable
M4 Max 128GB128GB163 tok/scomfortable
M4 Pro 48GB48GB81 tok/scomfortable
M4 Pro 24GB24GB81 tok/scomfortable
A100 PCIe 80 GB80GB818 tok/scomfortable
H100 SXM5 80 GB80GB1645 tok/scomfortable
GeForce RTX 309024GB394 tok/scomfortable
Radeon RX 7900 XTX24GB326 tok/scomfortable
GeForce RTX 408016GB316 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0