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

mdeberta-v3-base-squad2

by timpal0l

278M params · question-answering · 257 likes · 363.6k downloads

mdeberta-v3-base-squad2 is a 278M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (4719 tok/s), GeForce RTX 5090 (7077 tok/s), M4 Max 128GB (1726 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
HF Inference

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB4719 tok/scomfortable
GeForce RTX 509032GB7077 tok/scomfortable
M4 Max 128GB128GB1726 tok/scomfortable
M4 Pro 48GB48GB863 tok/scomfortable
M4 Pro 24GB24GB863 tok/scomfortable
A100 PCIe 80 GB80GB8646 tok/scomfortable
H100 SXM5 80 GB80GB17390 tok/scomfortable
GeForce RTX 309024GB4172 tok/scomfortable
Radeon RX 7900 XTX24GB3450 tok/scomfortable
GeForce RTX 408016GB3349 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0