Streaming 3D foundation model
LingBot-Map
Feed-forward 3D model for streaming video reconstruction; public KITTI/Oxford-Spires benchmarks.
Open-source code, paper, model releases, and public evaluation scriptsLocal open source
View demoOrganizationAnt Group / Robbyant
CategoryStreaming 3D foundation model
AvailabilityGitHub, arXiv, Hugging Face checkpoints, benchmark scripts in main repo.
Last reviewed2026-05-25
01 / Overview
Streaming 3D reconstruction, long-horizon spatial memory, pose estimation, browser-viewable scenes.
Feed-forward 3D model for streaming video reconstruction; public KITTI/Oxford-Spires benchmarks.
LingBot-Map matters here because it clarifies a specific lane: Streaming 3D reconstruction, long-horizon spatial memory, pose estimation, browser-viewable scenes.
This is a source-backed orientation profile, not a benchmark verdict.
02 / Strengths
Where it stands out
- Bridges world models and spatial computing via long-sequence geometry tracking.
- Grounds generated-3D-world coverage in a primary-source reconstruction system.
- Extends Ant coverage beyond LingBot-World into spatial, embodied AI.
- May 2026 release makes evaluation scripts and dataset preparation public.
03 / Boundaries
What not to overclaim
- Reconstruction model, not a generative world-creation product like Marble, HappyOyster.
- Focused on 3D perception and mapping, not general interactive simulation.
Primary evidence