← Back to digest
HF Trending Papers

TurboVLA: real-time VLA on a consumer GPU

What's this about?

A vision-language-action robot policy running at 32 Hz on an RTX 4090 with under 1 GB of VRAM — real-time control on consumer hardware.

Why it matters

Robot policies are usually heavy; hitting real-time on a single consumer GPU (9 upvotes) lowers the barrier to hands-on robotics research.

← Mage-VL: codec-native streaming multimodal modelMODUS: decoder-only any-to-any modeling →