All GPUs

RTX 4070 12 GB

NVIDIA · Ada Lovelace · 12 GB · 2023

Can this card run it?

Yes — run Qwen3.8-Flash-Next at Q2_0.

Download 70 GB · needs 48 GB system RAM.

Download the NVIDIA · CUDA build.

Specifications

VRAM12 GB
Memory bandwidth504 GB/s
Reading speed1376 tok/s
Writing speed36.9 tok/s
Board power350 W
Data sourcebenchmarked

These numbers come from our own benchmark runs on this card.

See the full benchmark