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

Olmo-3-7B-Instruct

by allenai

528K params · text-generation · 120 likes · 102.8k downloads

Olmo-3-7B-Instruct is a 528K parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (2484935 tok/s), GeForce RTX 5090 (3726456 tok/s), M4 Max 128GB (909338 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
Public AI

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB2484935 tok/scomfortable
GeForce RTX 509032GB3726456 tok/scomfortable
M4 Max 128GB128GB909338 tok/scomfortable
M4 Pro 48GB48GB454669 tok/scomfortable
M4 Pro 24GB24GB454669 tok/scomfortable
A100 PCIe 80 GB80GB4552749 tok/scomfortable
H100 SXM5 80 GB80GB9156976 tok/scomfortable
GeForce RTX 309024GB2197053 tok/scomfortable
Radeon RX 7900 XTX24GB1816860 tok/scomfortable
GeForce RTX 408016GB1763565 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0