LingBot-VA 2.0 Released
CyberRobooo · x · 2026-07-10
@robbyantbrain has released LingBot-VA 2.0, claiming it to be the first truly embodied-native foundation model. They emphasized that it was built from scratch for the physical world rather than being fine-tuned from a video generation model.
Key results and design specs include:
- RoboTwin 2.0 dual-arm task success rate of 93.6%
- Peak asynchronous execution frequency reaching 225Hz on an optimized FP8/TensorRT stack
- Strong generalization from just 10–15 demonstrations, with support for in-context learning using human reference videos without requiring parameter updates
Architecturally, it features a semantic vision-action tokenizer, a causal DiT with sparse MoE, and a mechanism called Foresight Reasoning that predicts the next world state during robot motion. The author also noted that multiple components released this week collectively form a "closed-loop" embodied full-stack:
- LingBot-Depth 2.0
- LingBot-Vision
- LingBot-VLA 2.0
- LingBot-World 2.0
- LingBot-Video
- LingBot-VA 2.0
The overarching claim is that a single "brain" can serve multiple robots.
Related event: LingBot-VA/VLA 2.0 Released: Native Embodied Foundation Model(24 posts)→
More from Embodied
- Full teardown: inside Zmotor's 32mm linear actuator for humanoids — MarwaEldiwiny · 2026-10-27
- Full Teardown of a Zmotor Linear Actuator Used in China's Humanoid Robots — MarwaEldiwiny · 2026-10-27
- World Model Assigns Weight and Friction to Objects, Then Simulates with a Physics Engine — jonstephens85 · 2026-10-27
- First unboxing shots of Apple's foldable "iPhone Duo" surface — chrisfirst · 2026-10-27
- Bedrock Robotics Raises $600M Series C as Autonomous Excavators Do Paid Work on US Job Sites — davidpattersonx · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27