LingBot-World 2.0: Open-Sourcing the Interactive World Model

Episode 1 · Ant-backed Team Open-Sources LingBot-World 2.0 Real-time World Model (2026-07-09, 26 posts)

An embodied AI team under Ant Group released and open-sourced LingBot-World 2.0 (Infinity), an interactive world model. The system aims to solve the common issue where AI world models quickly become blurry or distorted during movement, with the ultimate goal of creating "one brain for all robots."

Core Tech and Performance

LingBot-World 2.0 is an action-conditioned world model. By inputting a single frame, real-time camera poses, and text prompts, it autoregressively generates an explorable world. The team claims its long-term stability is a structural improvement rather than cherry-picking. In tests, it ran for 60 minutes across 20 different scenes without noticeable visual degradation, operating in real-time at 720p and 60fps. It also includes a Director Agent for advanced camera control, supporting fantasy scenes, racing cockpits, and cinematic visuals.

Interactive Experience and Architecture

The model turns AI-generated video into a "playable world." Users can walk using WASD keys, while the system pushes event options to hotkeys, allowing users to trigger attacks, archery, spellcasting, or summon storms and change weather. Architecturally, a VLM handles the next-step planning, a video generator handles rendering, and a pilot agent coordinates the process.

Open Source and Deployment

The release includes open-source model weights, inference code, and agentic-harness code on Hugging Face. The main model is 14B, but the paper also provides a 1.3B lightweight version that can run locally on a single consumer GPU. Additionally, the model is available to developers via the Reactor SDK and API for applications in gaming, creative tools, and simulations.

• LingBot-World 2.0 Released — _akhaliq · 2026-07-09
• LingBot-World 2.0 Focuses on Interactive Worlds — JaynitMakwana · 2026-07-09
• New Demo of Real-Time World Model — Scobleizer · 2026-07-09
• Lingbot-World-2 Demo: Playable Worlds — minchoi · 2026-07-09
• Interactive World Model Available for Download — OkCan8173 · 2026-07-09
• Interactive World Model Weights Released — Dramatic_Spirit_8436 · 2026-07-09
• Real-time Interactive World Model LingBot-World2.0 Open Sourced — 新智元 · 2026-07-09
• LingBot-World 2.0 Supports Single-GPU Deployment — testingcatalog · 2026-07-09
• Interactive World Generation Demo — eyishazyer · 2026-07-09
• Real-Time Controllable Open-Source World Model — Ill-Lengthiness1381 · 2026-07-09
• Lingbot-World-2 Launches Real-Time World Model — heyshrutimishra · 2026-07-09
• Open-Source Interactive World Model Playable in Real-Time — HarrySurden · 2026-07-09
• Real-Time World Model Lingbot-World-2 Launches — rohanpaul_ai · 2026-07-09
• LingBot-World 2.0 Released — thetripathi58 · 2026-07-10
• Ant's Embodied AI Team Unveils New World Model — thetripathi58 · 2026-07-10
• 720p60 Real-Time World Generation Gets More Stable — thetripathi58 · 2026-07-10
• Exploring an AI World Like a Video Game — thetripathi58 · 2026-07-10
• Open-Sourcing 14B World Model and 1.3B Version — thetripathi58 · 2026-07-10
• Try This World Model Directly Online — thetripathi58 · 2026-07-10
• LingBot-World 2.0 Open-Sourced — FinanceYF5 · 2026-07-10

6 more related posts →

Episode 2 · Ant Group Open-Sources LingBot-World 2.0 Interactive World Model (2026-07-11, 15 posts)

Robbyant, an embodied AI company under Ant Group, released LingBot-World 2.0 (also known as LingBot-World-Infinity), an open-source interactive causal world model. Moving beyond traditional video clip generation, this model pushes AI towards sustainably interactive "game worlds," drawing significant attention for its hour-level stable generation and multiplayer capabilities.

Core Tech and Performance

Utilizing a causal pre-training and post-training approach, the model specifically tackles the common autoregressive drift issue in world models (where accumulating errors distort the generated world). It achieved 720p/60fps real-time rendering over a continuous 60-minute rollout across 20 scenes, with almost no noticeable visual degradation. It requires only a single starting frame, combined with real-time camera movement or text input, to generate 3D environments.

Gameplay and Application Potential

Hailed as the first AI video model that plays like a real video game, LingBot-World 2.0 allows players to control characters using WASD keys, cast spells, and summon weather. Its interaction system uses a background VLM to suggest real-time events bound to keyboard hotkeys, categorized into fixed keys and context-aware events. Furthermore, it supports multiplayer, allowing multiple players to explore the same procedurally generated AI world simultaneously, likened to an AI version of Minecraft. Beyond entertainment, it holds potential for embodied simulations.

Open Source Availability

The project is fully open, with the author releasing model weights, GitHub code, an online demo, demonstration videos, and related papers.

• LingBot-World 2.0 Launches Interactive Video World Model — rohanpaul_ai · 2026-07-11
• LingBot-World 2.0 Released — rohanpaul_ai · 2026-07-11
• LingBot-World 2.0 Open-Sourced — CyberRobooo · 2026-07-11
• Hour-Scale World Models and Embodied Simulation — CyberRobooo · 2026-07-11
• LingBot-World-Infinity Focuses on Consistency — Shruti_0810 · 2026-07-11
• 7 Mainstream AI Image Models Reviewed — alifcoder · 2026-07-11
• Open-Source Real-Time AI World Runs for 60 Mins — heyshrutimishra · 2026-07-11
• LingBot 2.0: Open-Source World Model — kimmonismus · 2026-07-11
• Open-Source Interactive World Model LingBot-World 2.0 — FellMentKE · 2026-07-12
• LingBot World 2.0: Explore an AI-Generated World Like a Game — LearnWithBishal · 2026-07-12
• AI Generates an Entire Game World — LearnWithBishal · 2026-07-12
• LingBot World is Like a Playable AI Game — LearnWithBishal · 2026-07-12
• VLM Drives Real-Time Hotkey Interactions — LearnWithBishal · 2026-07-12
• Real-Time Multiplayer in AI-Generated World — LearnWithBishal · 2026-07-12
• Real-Time AI-Generated World Supports Multiplayer — LearnWithBishal · 2026-07-12