Standard vision models handle 3D tasks without custom architecture

Original: Excited to share VLM³ - standard VLMs go surprisingly far in 3D!

Source: x.com ↗