AMI Labs

AMI Labs develops world models intended to support memory, reasoning, and planning from real-world sensor data.

Research dossier · Reviewed 2026-09-21

Research and release status

The reviewed launch and mission pages describe a research program; they do not identify a public model release.

Predicting the world in representation space

AMI aims to learn abstract representations of sensor data and predict how actions change the world. Its stated approach discards unpredictable details so systems can plan. Industrial control, robotics, wearables, and healthcare are proposed applications, rather than evidence that each has been deployed. [1]

A distributed founding team

The official site names Alex LeBrun, Laurent Solly, Mike Rabbat, Min Lin, Pascale Fung, Saining Xie, and Yann LeCun. The March 10, 2026 launch announcement reports a $1.03 billion funding round and operations in Paris, New York, Montreal, and Singapore. [1][2]

What would establish progress

AMI commits to open publications and open source alongside work with industry partners. Research artifacts, evaluations of planning and memory, and concrete deployment reports will provide a firmer basis for assessment than the launch financing alone. [1][2]

Sources

[1] Mission and team ↗AMI Labs · accessed 2026-09-21
[2] Official launch ↗AMI Labs · accessed 2026-09-21