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

phi-1_5

by microsoft

1.4B params · text-generation · 1.4k likes · 144.4k downloads

phi-1_5 is a 1.4B parameter model. At Q4 quantization it requires 1GB of VRAM. It runs comfortably on GeForce RTX 4090 (925 tok/s), GeForce RTX 5090 (1388 tok/s), M4 Max 128GB (338 tok/s).

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB925 tok/scomfortable
GeForce RTX 509032GB1388 tok/scomfortable
M4 Max 128GB128GB338 tok/scomfortable
M4 Pro 48GB48GB169 tok/scomfortable
M4 Pro 24GB24GB169 tok/scomfortable
A100 PCIe 80 GB80GB1696 tok/scomfortable
H100 SXM5 80 GB80GB3411 tok/scomfortable
GeForce RTX 309024GB818 tok/scomfortable
Radeon RX 7900 XTX24GB676 tok/scomfortable
GeForce RTX 408016GB657 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0