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

LaBSE

by sentence-transformers

471M params · sentence-similarity · 331 likes · 1.2M downloads

LaBSE is a 471M parameter model. At Q4 quantization it requires 0GB of VRAM. It runs comfortably on GeForce RTX 4090 (2788 tok/s), GeForce RTX 5090 (4181 tok/s), M4 Max 128GB (1020 tok/s).

Inference providers

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

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB2788 tok/scomfortable
GeForce RTX 509032GB4181 tok/scomfortable
M4 Max 128GB128GB1020 tok/scomfortable
M4 Pro 48GB48GB510 tok/scomfortable
M4 Pro 24GB24GB510 tok/scomfortable
A100 PCIe 80 GB80GB5108 tok/scomfortable
H100 SXM5 80 GB80GB10274 tok/scomfortable
GeForce RTX 309024GB2465 tok/scomfortable
Radeon RX 7900 XTX24GB2038 tok/scomfortable
GeForce RTX 408016GB1978 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0