TwelveLabs Released Pegasus 1.6 for Physical AI
The new model enables machines to interpret egocentric video to improve robotic perception and learning.
Updated on Oct. 6, 2026 in Robotics
Live Poll
Do you trust that training AI on human tasks via video will make robots safer?
TwelveLabs has launched Pegasus 1.6, a new model specifically designed to help machines perceive and reason within physical environments. This technology allows robotics developers to train systems using human experience captured through video rather than starting from scratch.
Why it matters
By mapping human actions to domain-specific taxonomies, this model provides robotics teams with a way to build machines that can better interact with the real world. This capability is essential for advancing how AI agents handle physical tasks.
The Pegasus 1.6 model supports five core workflows, including action segmentation, dense captioning, and video clip quality scoring. It improves entity recognition by specifically tracking human hands, tools, and objects in egocentric video footage.
The players
TwelveLabs
TwelveLabs is a technology company based in San Francisco that focuses on video-native AI processing and physical world perception.
The details
The system processes egocentric footage to generate structured, reviewable knowledge, enabling robotics teams to accelerate training through observational data. Pegasus 1.6 also includes features for search, curation, and compliance flagging to ensure model performance meets safety standards.
Timeline
October 6, 2026: TwelveLabs officially released the Pegasus 1.6 model.
The Tech Race
This release follows the industry pattern of transitioning toward foundational models that leverage human visual data to improve physical world interaction. It positions TwelveLabs within the broader competition to bridge the gap between digital AI reasoning and physical task execution.
Developers using this technology gain access to improved tools for training robots, which could lead to more capable machines in manufacturing and logistics. These advancements in egocentric data processing help machines better identify and interact with household or professional tools.
The takeaway
The shift toward using egocentric video as a primary training source marks a significant step in making AI systems more reliable in physical settings. Developers should prioritize incorporating structured observational data to improve how machines interpret human-led workflows.
Further reading
For more information on the evolving landscape of autonomous systems, visit the Robotics section.
More information
Find complete technical specifications and use cases at the Pegasus 1.6 physical AI solution details.
Live Poll
Do you trust that training AI on human tasks via video will make robots safer?
