Meta's new AI model shows early promise, but investors are watching Zuckerberg's strategy

Meta unveiled a new AI model, Muse Spark, but investors are more focused on what strategy Mark Zuckerberg will present at the Q1 earnings call. Since this model was revealed at the start of Q2, explaining the future AI roadmap is expected to be the core of the earnings conference call.

Muse Spark is the first major model to come out of Meta Superintelligence Labs, led by Alexandr Wang. Wang is the former CEO of Scale AI, and Meta invested $14.3 billion in Scale AI to bring him on board. Zuckerberg subsequently reorganized the AI division further by also bringing on Nat Friedman and Daniel Gross.

The model's direction has also changed. Unlike the existing Llama, Muse Spark is a closed model that departs from the open-source, free-distribution path, and Meta is also reviewing plans to offer paid access to developers in the long term. While the company appears to be aiming for a direction of monetizing the model itself, similar to OpenAI, Anthropic, and Google, it is seen as still lacking a strategy to drive large-scale consumer usage—like ChatGPT or Claude—by attaching Meta AI to its core apps.

Meta's internal testing also showed weaker results than top-tier models, but JPMorgan nonetheless viewed Muse Spark as having brought Meta back into the AI competitive landscape. Based on Arena.AI, performance still lags behind the top tier.

• Text: behind Claude and Gemini, but ahead of GPT.
• Vision: behind only Claude.
• Documents/Code: Claude leads, and Meta ranks near the bottom.

Citizens assessed AI as a complementary good to Meta's advertising business, and Meta's Q1 revenue is expected to rise 31% year-over-year to $55.6 billion. The company presented 2026 AI-related CAPEX of $115 billion to $135 billion, and on May 20 it will cut about 8,000 jobs, or 10% of its total workforce.