Radar / AI robotics / Maxwell
Maxwell: embodied AI model tops Meta-World physical-task benchmark
Researchers at the Chinese Academy of Sciences' Institute of Artificial Intelligence for Industries reported Maxwell, an embodied AI model that scored 91.9 on the Meta-World robot manipulation benchmark - the highest ever recorded.
Why it matters
The first reported system to top every prior Meta-World submission, beating models from Physical Intelligence, Google DeepMind, and Alibaba - a data point that Chinese embodied models are leading on learning to act, not just to think.
What you could build with it
A robotics integrator could benchmark its own manipulation stack against Maxwell's Meta-World result to decide where learned policies beat hand-coded control; the concrete result is a factory or warehouse picking cell that adapts to new object layouts without reprogramming.
Does it hold up?
Too early to judge - only press coverage exists so far; no paper, model weights, or independent replication have surfaced, so the 91.9 figure rests on the Meta-World leaderboard submission alone.
Built with Maxwell
- NewsNation: Chinese AI model Maxwell tops physical-task benchmarknews · Second outlet covering the 91.9 score and the broader shift of Chinese research toward embodied AI for physical environments.
Learn more
First spotted on article: source.
More AI robotics
FLUX 3 ActionBlack Forest Labs launched FLUX 3 Action on Sept 24: a 7B open-weight world-action robotics model that takes camera…robotics · JEV 0.72Boston Dynamics fits Atlas humanoid with a factory-ready four-fingered handAtlas's new hand has 13 degrees of freedom (nearly double the prior three-fingered generation) with pressure sensors…robotics · JEV 0.66NVIDIA Isaac ROS 5.0: agentic AI workflows for roboticsIsaac ROS 5.0 moves to ROS 2 Lyrical/Ubuntu 24.04 and adds agent-ready skills - reusable workflows like…robotics · JEV 0.59Light-O1 (Light Origins): Embodied Foundation ModelGeneral-purpose embodied foundation model that learns a reusable human action prior from structured human actions…robotics · JEV 0.43Perceptron Mk1.5: perception model for robotics adds audio and explicit annotation outputsPerceptron shipped Mk1.5 (September 25), successor to its Mk1 vision-language model for video and embodied reasoning…robotics · JEV 0.38Isaac ROS 5.0GPU-accelerated robotics stack goes agentic with reusable AI agents for setup and manipulation tasks.robotics · JEV 0.32
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.