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

SmolLM2-1.7B-Instruct

by HuggingFaceTB

1.7B params · text-generation · 733 likes · 140.6k downloads

SmolLM2-1.7B-Instruct is a 1.7B parameter model. At Q4 quantization it requires 1GB of VRAM. It runs comfortably on GeForce RTX 4090 (767 tok/s), GeForce RTX 5090 (1150 tok/s), M4 Max 128GB (280 tok/s).

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB767 tok/scomfortable
GeForce RTX 509032GB1150 tok/scomfortable
M4 Max 128GB128GB280 tok/scomfortable
M4 Pro 48GB48GB140 tok/scomfortable
M4 Pro 24GB24GB140 tok/scomfortable
A100 PCIe 80 GB80GB1405 tok/scomfortable
H100 SXM5 80 GB80GB2827 tok/scomfortable
GeForce RTX 309024GB678 tok/scomfortable
Radeon RX 7900 XTX24GB560 tok/scomfortable
GeForce RTX 408016GB544 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0