Meta Platforms’ newly formed artificial intelligence research unit has reached an important internal milestone, with its first major AI models delivered inside the company this month, according to the company’s chief technology officer.
Speaking during a press interaction on the sidelines of the World Economic Forum’s annual gathering in Davos, Meta CTO Andrew Bosworth said the early results from the Meta Superintelligence Labs team looked highly encouraging. The lab was established last year as part of a broader restructuring of Meta’s AI strategy.
Bosworth noted that the team is still in the early stages of development, having worked on the models for less than six months so far. Despite the short timeline, he said the internal models demonstrated strong performance and showed clear potential, calling them “very good” in their current form. He did not confirm which specific models had been shared internally.
Industry chatter late last year suggested Meta was developing multiple AI systems, including a text-based model and another focused on image and video generation, both expected to play a role in the company’s next wave of AI products. Meta has not publicly confirmed timelines or features for these systems.
The company’s AI push is being closely watched after CEO Mark Zuckerberg reorganized Meta’s AI leadership, created the new lab, and aggressively recruited top talent with lucrative compensation packages. The moves came after Meta faced criticism over the performance of its Llama 4 model, particularly as competitors gained momentum in the fast-moving AI race.
Bosworth emphasized that training a model is only part of the journey. He said significant work is required after training to make AI systems reliable, practical, and usable for both internal teams and everyday consumers. Even so, he added that Meta is beginning to see positive results from its heavy investments made during what he described as a chaotic but necessary year of building infrastructure, securing computing power, and scaling its AI ambitions.
Looking ahead, Bosworth said 2026 and 2027 are likely to be defining years for consumer AI adoption. Recent advances, he explained, have already enabled models to handle common, everyday questions people ask at home, even though more complex tasks will continue to improve over time.
This shift makes the coming years critical for turning AI research into real consumer products. Meta is already moving in that direction, promoting its AI-enabled Ray-Ban smart glasses. The company recently paused their international rollout to focus on meeting strong demand in the U.S. market.
Also Read: OpenAI Acquires Torch to Enhance ChatGPT’s Health Features
