News
Meta Unveils Muse Spark 1.1, Expands Push into Enterprise AI
Meta is stepping up its enterprise AI ambitions with the launch of Muse Spark 1.1, a multimodal reasoning model designed for agentic AI, alongside a new Meta Model API that gives developers access to the model for the first time.
Announced in a statement on July 9, Muse Spark 1.1 succeeds the company's original Muse Spark model with improvements in coding, tool use, computer interaction, and multimodal reasoning.
The launch comes as competition among foundation model providers shifts from building larger models to attracting developers. OpenAI, Anthropic, and Google have all expanded their APIs and agentic AI offerings, making developer platforms a key battleground for enterprise AI. At the same time, AI vendors are increasingly competing not only on model capability, but also on pricing, deployment, and developer ecosystems as enterprises move AI applications into production.
According to Meta, Muse Spark 1.1 is designed to plan, reason, and complete multi-step tasks across multiple applications with minimal human intervention.
The company said the model can coordinate multiple AI agents, maintain context across long workflows, and support a 1-million-token context window. Meta also announced a public preview of the Meta Model API, pricing Muse Spark 1.1 at $1.25 per million input tokens and $4.25 per million output tokens, positioning the model aggressively against competing frontier AI offerings.
Meta also highlighted improvements in coding and computer use, positioning Muse Spark 1.1 for software engineering and agentic development workflows. The California-based company said the model can diagnose bugs, implement features, perform large-scale code migrations, and determine when to automate tasks with scripts rather than relying solely on user interface interactions.
According to Meta, it conducted extensive safety testing under its Advanced AI Scaling Framework, with Muse Spark 1.1 demonstrating improved resistance to jailbreaks, prompt injection attacks, and hallucinations.
Muse Spark 1.1 is the flagship model from Meta Superintelligence Labs, the company's AI organization focused on developing frontier AI systems for both consumer and enterprise applications.
For developers, the biggest announcement may be the Meta Model API. By opening Muse Spark 1.1 to third parties, Meta is positioning itself as a platform for building AI agents and enterprise applications.
The move places Meta in more direct competition with developer platforms from OpenAI, Anthropic, and Google, which have all expanded APIs and enterprise AI offerings over the past year. This reflects a broader industry shift in which success depends not only on model performance, but also on attracting developers and becoming the platform on which enterprise AI applications are built.
The move also marks an evolution in Meta's AI strategy. While the company has focused heavily on consumer AI through Meta AI and its family of apps, the new API positions Meta more directly alongside companies competing to become the infrastructure layer for enterprise AI. It also reflects Meta's effort to pair its vast consumer AI ecosystem with a growing enterprise developer platform.
"We're thrilled to be releasing Muse Spark 1.1, a testament to our research momentum," Meta said. "We have even more capable models in training and look forward to sharing what's to come."
The launch reflects a broader shift across the AI industry, where competition is expanding beyond benchmark performance to include pricing, developer adoption, and the ability to deploy AI efficiently at enterprise scale. As organizations build more agentic AI applications, vendors are increasingly competing on the overall economics of AI platforms as much as on the capabilities of their underlying models.