Mini toolsLingBot VLA 2.0
By Robbyant
Model family: LingBot
LingBot VLA 2.0 is a Vision Language Action foundation model built to move from large scale pretraining to reliable real world robot deployment. It is trained on about 60,000 hours of data, including 50,000 hours of robot trajectories across 20 robot configurations and 10,000 hours of egocentric human video. It uses a unified 55 dimensional action representation covering arms, end effectors, grippers, dexterous hands, waist, head, and mobile base signals, supporting single arm, dual arm, humanoid, and mobile manipulation embodiments. A sparse mixture of experts action expert improves cross embodiment scaling, and dual query distillation from depth and video teacher models adds current and future scene understanding. On an RTX 4090D, inference takes about 130ms with 10 denoising steps. Released under Apache 2.0.
New
Multimodal
Overview
LingBot VLA 2.0 is a 6B parameter Vision Language Action foundation model that converts camera images and language instructions into robot actions across arms, grippers, dexterous hands, waist, head, and mobile base signals for real world manipulation and mobile robot tasks.
Last updated: July 8, 2026
