← All models

Apollo World

Published Publisher Baidu

Overview

Apollo World is an agent-driven, multimodal world model for physical AI, focused on simulating environments for autonomous-driving training and validation. Baidu first disclosed it on November 3, 2026, at its AI Day · Apollo event in Shenzhen. (sohu.com)

Capabilities and uses

According to Baidu, the model represents vehicles, pedestrians and other road users as interactive agents that respond to the autonomous vehicle’s actions. It generates camera images and LiDAR point clouds, updates the simulated environment in response to driving decisions, and supports a closed loop of perception, decision-making, action and feedback. It can simulate complete trips through a virtual city, from pickup to destination, and turn real-world complex or risky situations into repeatable tests. Baidu says it is used in Apollo Go’s global testing and validation. (sohu.com)

Baidu also proposed a 2+3 evaluation framework: causality and interactivity as foundational requirements, plus physical accuracy, persistence and real-time performance as advanced criteria. These are Baidu’s proposed criteria, not independently verified performance results. (sohu.com)

Limitations and access

Simulation results do not by themselves establish real-world driving safety; Baidu describes subsequent testing of updated driving systems on real roads. The announcement does not specify a public API, model ID, parameter count, model-weight availability, or independent benchmark results. (sohu.com)

Pricing

As of November 18, 2026, no official Apollo World pricing could be verified on Baidu’s domestic or international sites. CNY and USD rates, including any input/output charges or other billing units, are unverified; no public pricing is assumed.

Official source

Baidu’s announcement on Sohu, November 3, 2026. (sohu.com)