Meta's new model is Muse Spark, and meta.ai chat has some interesting tools

# Meta Announces Muse Spark Model with Limited Availability

Meta has released Muse Spark, marking its first model launch since Llama 4 approximately one year ago. Unlike Meta's previous open-source approach, Muse Spark is a hosted model rather than open-weights, meaning users cannot access or modify the underlying code. The model is currently available through a limited private API preview for select users, though the general public can experiment with it through Meta's meta.ai platform, which requires a Facebook or Instagram login.

The release represents a strategic shift in Meta's AI deployment model. By offering Muse Spark as a hosted service rather than open-source software, Meta maintains greater control over the model's usage and monetization while reducing the barrier to entry for casual users through the web interface. This hybrid approach—private API for developers and public access through meta.ai—allows Meta to manage resource allocation and gather user feedback before broader rollout.

The timing and limited availability suggest Meta is taking a measured approach to Muse Spark's launch, likely to assess performance, manage server costs, and refine the model based on early user feedback. The requirement for social media login ties the model's usage to Meta's broader ecosystem, positioning it as a competitive offering in the crowded generative AI market alongside competitors like OpenAI's ChatGPT and Google's Gemini.