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

StableBeluga2

by stabilityai

78.4B params · text-generation · 882 likes · 1.1k downloads

StableBeluga2 is a 78.4B parameter model. At Q4 quantization it requires 39GB of VRAM. It runs comfortably on A100 PCIe 80 GB (30 tok/s), H100 SXM5 80 GB (61 tok/s), H200 SXM 141 GB (89 tok/s).

GPU compatibility

GPUVRAMQ4 DecodeVerdict
M4 Max 128GB128GB6 tok/stight
M4 Pro 48GB48GB3 tok/stight
A100 PCIe 80 GB80GB30 tok/scomfortable
H100 SXM5 80 GB80GB61 tok/scomfortable
M4 Max 64GB64GB6 tok/stight
H200 SXM 141 GB141GB89 tok/scomfortable
M2 Ultra 192GB192GB8 tok/stight
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0