Meet Kyutai REKI: The 2.4B VLA Model Bringing Efficient Memory to Robot Manipulation
The nonprofit research lab Kyutai has introduced REKI, a 2.4-billion parameter vision-language-action model designed specifically for robotic manipulation. By utilizing a compact memory mechanism that retains recent operational history, REKI achieves competitive performance against much larger models while relying exclusively on public training data. This development is a milestone for lightweight robotics, proving that sophisticated spatial reasoning and task execution can be achieved with significantly lower computational overhead and energy requirements.
What this means for you
Developers and robotics engineers should explore REKI to build high-performance, resource-efficient automation systems without relying on massive, proprietary datasets.
The linked source has not been established as the original announcement.
Put this to work
Try the tools in this story inside the AI Tool Lab, browse roles in the AI Job Portal, test your skills with Aptitude Tests, or read our in-depth guides.
Keep reading
More AI news
Meta removes language that raised scraping and AI training concerns
Meta has revised its terms of service to address ambiguity regarding how user data is utilized for AI model training. While the company maintains that its web-crawling practices remain unchanged, the update clarifies language that previously sparked significant privacy concerns among users and regulators. By refining these terms, Meta aims to mitigate potential legal friction and demonstrate greater transparency, ensuring that its ongoing development of generative AI tools remains within the bounds of evolving global data protection standards.
Source check and human review not documented.
Linked source: TechCrunch AI (original announcement not established)Read the briefAnthropic says its workforce has become less junior and is spending more time in meetings
Anthropic has reported a notable shift in its organizational structure over the past year, transitioning from a primarily junior-heavy startup to an enterprise with a more experienced workforce. As the company scales, internal data suggests a rise in collaborative overhead, with employees dedicating significantly more hours to meetings compared to individual coding or research tasks. This evolution highlights the growing pains common to AI labs transitioning into established corporate entities while managing complex, high-stakes model deployments.
Source check and human review not documented.
Linked source: TechCrunch AI (original announcement not established)Read the briefReplit is no longer working with Anthropic, CEO says
Software development platform Replit has officially selected OpenAI as its primary artificial intelligence provider, stepping away from its previous default partner, Anthropic. Replit CEO Amjad Masad announced the strategic shift, noting that while OpenAI's models will now power the core coding experience by default, Anthropic's Claude models will remain accessible to users who prefer them. This transition highlights intensifying competition among major AI providers to secure deeply integrated developer ecosystems, signaling how shifting performance benchmarks and commercial partnerships influence platform-level software tooling choices.
Source check and human review not documented.
Linked source: TechCrunch AI (original announcement not established)Read the brief