Meta Re-enters AI Race with New Multimodal Model

Meta has officially entered the advanced AI model race with the introduction of Spark Muse, a natively multimodal reasoning model. This new AI system, developed by Meta Superintelligence Labs, represents a substantial upgrade over the company's previous AI efforts.

Spark Muse is engineered with capabilities such as tool-use, visual chain-of-thought processing, and multi-agent orchestration. It is poised for integration into Meta's flagship products, including Facebook, Instagram, Messenger, and WhatsApp, with a rollout anticipated in the upcoming weeks.

According to Meta's benchmarks, Spark Muse demonstrates strong performance, ranking behind only Google's Gemini 3.1 Pro and OpenAI's GPT-5.4 in multimodal functionality. It also shows competitive results in reasoning tasks, though it trails behind leading models from Anthropic and Google.

The model exhibits particular strengths in its capacity to draw styling inspiration from user-followed creators and communities to offer personalized shopping recommendations. Furthermore, Meta highlights its ability to process health data, a common consumer use case for AI applications.

While Spark Muse may not drastically alter the current AI landscape, its release positions Meta to compete more effectively. The company aims to build upon this new baseline for future advancements, moving Meta from being out of the race to a contender for a significant position.