Meta Launches Muse Spark 1.1 To Take On OpenAI And Anthropic

(RTTNews) - Meta Platforms Inc. (META) has publicly rolled out Muse Spark 1.1, an upgraded multimodal AI model aimed at agentic coding and software development part of the company's push to compete more directly with OpenAI and Anthropic.

Announced by Meta AI chief Alexandr Wang, the model is now in public preview through Meta's developer portal. Some early partners already have access, while other developers can join a waitlist to use the API. For now, the service will run only on Meta's own infrastructure, rather than on third-party platforms.

Wang called Muse Spark 1.1 Meta's strongest model so far for agentic and coding work. He said it can handle multistep reasoning, manage complex workflows, fix bugs, and support large-scale code migrations. The model was trained to integrate with widely used developer tools and AI frameworks so autonomous agents can take on more sophisticated tasks.

Meta is also positioning the API as a lower-cost option. New users get $20 in free credits, after which pricing is $1.25 per million input tokens and $4.25 per million output tokens. Wang said the goal is to drive adoption through aggressive pricing.

The launch follows Meta's release of Muse Image earlier this week and highlights the company's broader effort to monetise its growing AI investment. While Meta has long leaned into open-source AI through its Llama models, it has increasingly been bringing proprietary models to market. Wang said an open-source version of Muse Spark is still in development, but he didn't share a timeline.

Meta is also reportedly working on a more advanced model, code-named Watermelon, as it looks to strengthen its position in a fast-moving AI landscape.