Breaking

AGIBOT open-sources Genie Sim 3.0, an LLM-driven robot simulator, at CES 2026

July 1, 2026 at 03:59 EDT

Robotics
Open Source
AI Agents

Chinese robotics company AGIBOT has unveiled and open-sourced Genie Sim 3.0 at CES 2026, a simulation platform that uses LLMs to automatically generate synthetic scenes and evaluate embodied AI. The platform generates high-fidelity 3D scenes from natural-language instructions in minutes and ships with a synthetic dataset exceeding 10,000 hours, releasing code, assets and data in full. The platform is integrated with NVIDIA Isaac Sim (built on Omniverse) and is available on GitHub and ModelScope.

CES 2026 · AGIBOT

Genie Sim 3.0: an open-source simulation stack for embodied AI

A fully open-source platform that turns natural-language prompts into high-fidelity 3D robot scenes — unifying environment reconstruction, data collection, training and automated evaluation into a single toolchain.

10,000+ h

synthetic data (200+ tasks)

100,000+

evaluation scenarios in the benchmark

5,140

assets · 353 categories · 5 domains

<10%

sim-to-real gap

Simulation-trained vs real-data baseline

Average task success rate — synthetic data can outperform a real-data baseline.

0.75

Real-data
baseline

0.83

Simulation-
trained

The full-stack loop

Environment

Real capture → 3D reconstruction (MetaCam, LiDAR, RGB)

→

Data

LLM scene generation + domain randomization

→

Training

VLA models on synthetic data

→

Evaluation

VLM-based automated assessment

Evaluation axes: Instruction · Spatial · Manipulation · Robustness · Sim2Real

Praised for

LLM-driven 3D environment generation and full-stack integration. At the AGIBOT WORLD CHALLENGE 2026, 526 teams used it as a framework linking through to real-world validation.

Worth noting

Detailed user experiences and reported limitations remain scarce — an early-stage project whose practical value will be tested as issues and future challenge results accumulate.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.