Back
Meta launches paid Muse Spark 1.1 API as it weighs AI cloud services

Meta launches paid Muse Spark 1.1 API as it weighs AI cloud services

Muse Spark 1.1 debuts via Meta Model API with a 1 million-token context window, low-cost pricing, and early partners including Replit, Cline, and Box, as Meta also explores selling model access and compute capacity.

Fact Check
The claim is strongly corroborated. Zuckerberg's own Threads post confirms Meta released Muse Spark 1.1 through a 'new Meta Model API,' and both the Reuters and CNBC articles confirm it is Meta's first paid, pay-as-you-go developer model, marking a shift from free open-source Llama to proprietary paid access. The cryptobriefing/Bloomberg-derived report confirms Zuckerberg is actively considering an AI cloud business—renting out compute and API access. All elements of the claim (first paid developer model, API access monetization, and Zuckerberg exploring selling compute/cloud) are supported by these sources.
Summary

Meta has launched Muse Spark 1.1, its first paid AI model for developers, through the Meta Model API and Meta AI platform, marking a shift beyond mostly open or free AI offerings. Released on July 9, 2026 by Meta Superintelligence Labs, the upgraded multimodal reasoning model is built for agentic coding and broader tool-using tasks, includes a 1 million-token context window, supports sub-agents, and is priced at $1.25 per million input tokens and $4.25 per million output tokens, with $20 in free credits for new users. Meta says the model can use software and tools, operate a computer across desktop, mobile and browser environments, and handle images and video as well as text. The launch comes as Mark Zuckerberg says Meta is also exploring an AI cloud business and whether leasing some compute capacity or selling model-hosting services to outside customers could create more value.

Terms & Concepts
  • Meta Model API: Meta’s interface for outside developers to access and integrate its AI models into their own applications.
  • agentic coding: An AI approach in which models carry out multi-step software tasks such as using tools, diagnosing bugs, and implementing features.
  • context window: The amount of text, code, or other input an AI model can retain and use in a single working session.