Local open source · Open stack · Company profile

Ant Group / Robbyant

Ant Group's Robbyant stack spans LingBot-World, LingBot-Map, LingBot-VA, and LingBot-VLA across simulation, reconstruction, and embodied control.

Official source

What this company is for

Robbyant connects world models with maps and robot control.

Consumer access. Local open source

Visitor can see. Open repos, papers, model releases spanning worlds, maps, robot control.

Where the claim stops. Good category background; not a consumer world creator.

This company's models

Models from Ant Group / Robbyant

Open-source interactive world simulator

LingBot-World

Robbyant's open-source world simulator for high-fidelity, long-horizon, real-time interactive environments.

GitHub, arXiv report, Hugging Face model cards, ModelScope downloads.

Serious GPUs required; mobile experience is product-mediated, not local inference.

Open the LingBot-World dossier

Streaming 3D foundation model

LingBot-Map

Feed-forward 3D model for streaming video reconstruction; public KITTI/Oxford-Spires benchmarks.

GitHub, arXiv, Hugging Face checkpoints, benchmark scripts in main repo.

Reconstruction model, not a generative world-creation product like Marble, HappyOyster.

Open the LingBot-Map dossier

Robot-control world model

LingBot-VA

Robbyant's causal video-action world model for generalist robot control.

GitHub, arXiv paper, project page, Hugging Face model releases.

No consumer explorable world product like HappyOyster, Marble, Genie 3.

Open the LingBot-VA dossier

Vision-language-action foundation model

LingBot-VLA

Robbyant's pragmatic vision-language-action model for generalist robotic manipulation across platforms.

GitHub, arXiv report, Hugging Face cards and collection, ModelScope checkpoints.

No consumer world-building product; not HappyOyster, Marble, or Genie 3.

Open the LingBot-VLA dossier

What to open first

Robbyant is a stack, not one demo.

LingBot-World is the interactive simulator. LingBot-Map is streaming 3D reconstruction. LingBot-VA is a causal video-action world model for robot control. LingBot-VLA is a vision-language-action policy family.

Opening only one repo will make Ant look narrower than the public evidence. Opening all four as the same product will overclaim. The company page exists to keep those lanes labeled.

LingBot-WorldOpen interactive world simulator.
VA vs WorldRobot-control model versus simulator.

Access boundary

Open weights are not the same as a consumer world app.

GitHub, arXiv, Hugging Face, and ModelScope are the inspectable surfaces. Runtime cost is real: serious GPUs are the default assumption for local use.

Mobile or app-mediated experiences, including Lingguang coverage, should stay separate from the claim that a reader can run the open model on a phone.

How to verify

Match the question to the repo before citing Ant as one system.

Ask whether the job is simulation, mapping, action-conditioned prediction, or instruction-following manipulation. Then open the matching LingBot dossier and its paper or model card.

If the reader wanted a no-code world creator, send them to Marble or HappyOyster instead of stretching Robbyant into that lane.

Keep reading

open-source world model stacksUse Tencent Hunyuan, Robbyant, Oasis, and NVIDIA Cosmos when the job is code, weights, papers, or evaluation, not a consumer editor.
physical AI world models guideRead NVIDIA Cosmos, Robbyant LingBot, and Tencent HY-Embodied as physical-AI lanes. They are not creator worlds, and they are not interchangeable robot products.
LingBot-VA vs LingBot-WorldRobot-control world modeling or an open explorable-world simulator?
LingBot-VLA vs LingBot-VARobot policy or world model predicting visual dynamics and actions?
LingBot-Map vs MarbleReconstructing spatial geometry, or generating a new editable 3D world?
HappyOyster vs LingBot-WorldEarly-access creation product or open-source simulator stack?
Ant Lingguang brings a world-model experience to mobileLingguang reportedly adds 'Experience World Model' backed by LingBot-World-Fast.
Robbyant releases LingBot-Map for streaming 3D reconstructionLingBot-Map adds public evaluation scripts for KITTI and Oxford Spires.
Robbyant publishes LingBot-VA for robot-control world modelingLingBot-VA: causal video-action world model predicting dynamics and actions together.
Robbyant releases LingBot-VLA for embodied-AI controlLingBot-VLA: vision-language-action model for robot manipulation, with public checkpoints.