Meta Platforms is sending mixed signals about its artificial intelligence progress. On one hand, Alexandr Wang, the company’s superintelligence chief, told employees that the upcoming Watermelon AI model has caught up with OpenAI’s GPT-5.5 — a claim that would mark a major internal milestone. On the other, CEO Mark Zuckerberg acknowledged in the same town hall that Meta’s AI agent technology has not advanced as quickly as expected, citing management and product challenges.
Wang said Watermelon uses “an order of magnitude” more compute than the Avocado model (the internal codename for Muse Spark, released in April). He also teased an imminent Muse Spark update with “major gains” in coding and agentic capabilities, replying on X that a model competitive with Anthropic’s Claude Opus would be coming “pretty soon.” However, neither Meta nor OpenAI provided independent verification of the GPT-5.5 parity claim.
Despite these optimistic signals, Meta’s stock fell 4.90% on Friday. The drop came as Zuckerberg admitted the company’s restructuring — which reassigned roughly 7,000 employees to AI teams in May and included significant job cuts — had not been “as clean as it could have been.” Reuters reported that the reset triggered employee pushback and morale concerns, adding pressure to the AI-driven strategy. Meta’s 2026 infrastructure spending forecast also rose to between $125 billion and $145 billion, up from a prior $115–$135 billion estimate, driven by rising component and data center costs.
OpenAI released GPT-5.5 in April and followed up with GPT-5.6 late last month, though the latter is not yet broadly available. Wang’s claims and Zuckerberg’s admission leave the near-term trajectory of Meta’s AI ambitions uncertain, especially as the company seeks significant benefits from its AI investments within three to six months.
