Maxwell Scores 91.9 on the Meta-World Robotics Benchmark
Published by AINave Editorial
Maxwell, an embodied AI model developed by the Chinese Academy of Sciences’ Institute of Artificial Intelligence for Industries, scored 91.9 on Meta-World, which the South China Morning Post describes as the benchmark’s highest recorded score. That puts Maxwell at the top of this particular simulated robotics test, not on a general leaderboard of robot capability. The reported score and development team
What the Meta-World result measures
Meta-World was set up by researchers from Stanford University, UC Berkeley and other institutions to evaluate robot performance across 50 everyday physical tasks. The examples include grasping and carrying objects, opening doors and using drawers. The benchmark’s task set
That range makes the score a useful signal about performance across simulated manipulation tasks. But the available reporting does not give further scoring-method details, so 91.9 should be read as the reported benchmark result, not converted into a broader claim about reliability, generalization or readiness for physical robots.
A notable lead, without a score-by-score comparison
The benchmark’s submissions include teams from Physical Intelligence, Google DeepMind, Alibaba, Meituan, Carnegie Mellon University, MIT, the University of Cambridge and the Toyota Research Institute. The report does not provide those entries’ scores, so it establishes Maxwell’s reported top position but not its margin over any specific competitor. The listed participating teams
The South China Morning Post also discloses that Alibaba owns the publication, a relevant detail because Alibaba is among the teams named in the benchmark field. The ownership disclosure and team list
Simulation is not a deployment result
Meta-World evaluates tasks in simulation. The report does not specify tests or outcomes for Maxwell on physical robots, so its benchmark lead alone cannot establish how the model would handle the same tasks in a real environment. The benchmark description and reported score
For robotics teams, the useful conclusion is narrow but meaningful: Maxwell has a reported top score on a benchmark built around varied everyday tasks. Whether that performance carries over from simulation to physical robots remains unanswered by the available reporting.
