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

Qwen2.5-Coder-3B-Instruct

by Qwen

3.1B params · text-generation · 96 likes · 125.4k downloads

Qwen2.5-Coder-3B-Instruct is a 3.1B parameter model. At Q4 quantization it requires 2GB of VRAM. It runs comfortably on GeForce RTX 4090 (425 tok/s), GeForce RTX 5090 (638 tok/s), M4 Max 128GB (155 tok/s).

Inference providers

Provider$/1M in$/1M outThroughput
Nscale77 tok/s

GPU compatibility

GPUVRAMQ4 DecodeVerdict
GeForce RTX 409024GB425 tok/scomfortable
GeForce RTX 509032GB638 tok/scomfortable
M4 Max 128GB128GB155 tok/scomfortable
M4 Pro 48GB48GB77 tok/scomfortable
M4 Pro 24GB24GB77 tok/scomfortable
A100 PCIe 80 GB80GB779 tok/scomfortable
H100 SXM5 80 GB80GB1567 tok/scomfortable
GeForce RTX 309024GB376 tok/scomfortable
Radeon RX 7900 XTX24GB311 tok/scomfortable
GeForce RTX 408016GB301 tok/scomfortable
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0