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

Qwen2.5-VL-72B-Instruct

by Qwen

73.4B params · image-text-to-text · 629 likes · 514.9k downloads

Qwen2.5-VL-72B-Instruct is a 73.4B parameter model. At Q4 quantization it requires 37GB of VRAM. It runs comfortably on A100 PCIe 80 GB (32 tok/s), H100 SXM5 80 GB (65 tok/s), H200 SXM 141 GB (95 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
Featherless
OVHcloud36 tok/s

GPU compatibility

GPUVRAMQ4 DecodeVerdict
M4 Max 128GB128GB6 tok/stight
M4 Pro 48GB48GB3 tok/stight
A100 PCIe 80 GB80GB32 tok/scomfortable
H100 SXM5 80 GB80GB65 tok/scomfortable
M4 Max 64GB64GB6 tok/stight
H200 SXM 141 GB141GB95 tok/scomfortable
M2 Ultra 192GB192GB9 tok/stight
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0