Meta has introduced Muse Spark 1.3 for coding and agentic workflows, emphasizing multi-step execution and long-running engineering tasks. The most consequential part of the release is its two-tier pricing: users who allow their activity to contribute to model improvement receive a substantially lower rate, while privacy-sensitive teams pay the standard price.
That turns model selection into a data-governance decision rather than a simple benchmark comparison. Meta has also signaled an open-weights release for the model family, which could give developers a path between a managed API and deployment inside their own controlled environment.
Price competition is turning data-use rights into an explicit cost that companies must measure and approve.