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

stable-code-3b

by stabilityai

2.8B params · text-generation · 659 likes · 6.4k downloads

stable-code-3b is a 2.8B parameter model. At Q4 quantization it requires 1GB of VRAM. It runs comfortably on GeForce RTX 4090 (469 tok/s), GeForce RTX 5090 (704 tok/s), M4 Max 128GB (171 tok/s).

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB469 tok/scomfortable
GeForce RTX 509032GB704 tok/scomfortable
M4 Max 128GB128GB171 tok/scomfortable
M4 Pro 48GB48GB85 tok/scomfortable
M4 Pro 24GB24GB85 tok/scomfortable
A100 PCIe 80 GB80GB860 tok/scomfortable
H100 SXM5 80 GB80GB1730 tok/scomfortable
GeForce RTX 309024GB415 tok/scomfortable
Radeon RX 7900 XTX24GB343 tok/scomfortable
GeForce RTX 408016GB333 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0