Figure has announced Helix, an artificial intelligence developed for integration with robots. According to the creators, the model is capable of “reasoning like a human.”

Equipped with AI humanoids, they can pick up almost any household item without the need for prior training or programming.

Today, we're launching our first technical report!

The report details Helix, a generalist Vision-Language-Action (VLA) modelhttps://t.co/Abl8qJDfVA
— Figure (@Figure_robot) February 20, 2025

“When asked to “pick up an item from the desert,” Helix identifies the toy cactus, selects the closest hand, and performs precise movements to securely grab it,” says the Figure team.

According to the technical report, the new Vision-Language-Action universal model enables machines to interpret natural language commands and interact with unfamiliar objects without the need for prior programming.

As a result, robots become smarter over time without system updates. In a demonstration, the robot was unaware of the items it was working with, but was able to determine which ones to place in the fridge.

Helix runs on energy-efficient embedded graphics processors, making the system suitable for commercial use. Figure has already signed contracts with BMW Manufacturing and a unnamed major U.S. client.

Reminder: In February, Figure broke its agreement with OpenAI, making a “significant breakthrough” in the development of AI for humanoids.