top of page
Search

Meta hires ChatGPT co-creator to lead its new Superintelligence Lab


ree

Meta hires ChatGPT co-creator Shengjia Zhao to lead its new Superintelligence Lab. The former GPT-4 architect will guide Meta's next-generation AI research in direct collaboration with Zuckerberg and Alexandr Wang.


Meta has officially appointed Shengjia Zhao, one of the original co-creators of ChatGPT and a key architect behind GPT-4 and the o-series models, as Chief Scientist of Meta Superintelligence Labs. The announcement, made public on July 25, 2025, marks one of Meta’s boldest AI hires to date. Zhao will oversee the design and training of Meta’s most advanced models, including future versions of LLaMA, and report directly to Chief AI Officer Alexandr Wang, with close collaboration from CEO Mark Zuckerberg.



The move underscores Meta’s ambition to regain ground in the AI space by recruiting elite talent from OpenAI and other competitors, consolidating leadership around a newly formed lab focused exclusively on superintelligence research.


Zhao will lead long-term AI research focused on reasoning, autonomy, and scalability

The Meta Superintelligence Lab (MSL) is distinct from Meta’s FAIR division and was created to push the boundaries of what Meta calls “scalable intelligence systems.” Shengjia Zhao, known in the industry as a pioneer of GPT-4’s reasoning engine and early versions of the “o3” model family, will now shape Meta’s most experimental AI initiatives.


His mandate includes:

  • Building multimodal reasoning architectures to rival GPT-5 and Claude Opus

  • Developing agentic frameworks designed for autonomous task execution

  • Guiding the next iterations of LLaMA, including distributed training optimization

  • Leading recruitment of top-tier researchers into Meta’s core AI teams


Zhao’s hiring signals Meta’s intent to move beyond assistive tools and pursue general-purpose reasoning agents trained on multi-trillion token datasets and supported by custom-built AI infrastructure.



Meta backs its AI strategy with massive infrastructure and talent incentives

Zhao’s arrival comes amid Meta’s largest infrastructure expansion to date. The company is building a new generation of “promethium-scale” data centers, equipped with next-gen GPUs and optimized for sustained training of frontier models. According to internal estimates, Meta’s AI budget will reach $65–70 billion annually by 2026.


To attract researchers of Zhao’s caliber, Meta has reportedly offered:

  • Signing packages of $80–100 million

  • Rapid lab build-outs in Menlo Park and London

  • Research autonomy with direct access to Meta’s compute clusters

  • A long-term roadmap tied to both open-source and proprietary AI releases



These investments place Meta in direct competition with OpenAI, Anthropic, and Google DeepMind not only on model performance—but on recruitment, publishing rights, and real-world product timelines.


The team now includes several former OpenAI and Google scientists

With Zhao now in place as Chief Scientist, Meta’s Superintelligence Lab has added at least a dozen researchers previously affiliated with OpenAI’s reasoning and infrastructure groups. Among them are:

  • Camille Chen, formerly of OpenAI’s alignment team

  • Victor Hsu, expert in high-bandwidth training optimization

  • Alessandra Jung, ex-Google Brain, specialized in data curation at scale


This mix of former OpenAI, DeepMind, and Anthropic talent is part of a broader trend: Meta is quietly becoming a gravitational hub for senior AI researchers looking for both compute access and strategic autonomy.



What to expect next from Meta’s Superintelligence team

The first wave of outputs under Zhao’s leadership is expected in late 2025, possibly aligned with an upgraded LLaMA 4.5 model and the internal preview of Meta’s new AI agentic system. Rumors within the industry suggest Meta is also preparing:

  • A new open-source benchmark suite focused on reasoning under uncertainty

  • A self-hosted inference engine for “always-on” edge deployment

  • Integrations between Meta AI models and WhatsApp Business Tools


With this appointment, Meta confirms it is no longer simply chasing the lead in AI—it’s aiming to define the next architecture era, starting from its own research lab and guided by the same minds that helped build GPT itself.


__________

FOLLOW US FOR MORE.


DATA STUDIOS

bottom of page