Meta launched Muse Spark 1.1 on Thursday, a major update to its AI model aimed at competing with OpenAI and Anthropic in coding and agentic workloads. It is the company’s strongest coding-and-agentic model yet, according to AI chief Alexandr Wang, and is being offered through a developer portal public preview with sign-up access and integration instructions.<\/p>\n
Three months after unveiling its first artificial intelligence model under AI chief Alexandr Wang, Meta is rolling out a significant update designed to compete with OpenAI and Anthropic in AI coding and agentic workloads.<\/p>\n
The initial Muse Spark model, released in April, was limited to select partners via a private API preview. The new version is being made available through a developer portal as part of a public preview, where users can sign up and view integration instructions. A Meta spokesperson said some early partners already have access to the API, and new users can join a waitlist and be added over time.<\/p>\n
For now, Meta is restricting API access to its own properties rather than distributing the model through third-party platforms such as the OpenRouter marketplace. Wang described the rollout as “served on top of the computer infrastructure that we’ve built.”<\/p>\n
Wang characterized the pricing for Muse Spark 1.1 as “very aggressive and attractive” relative to comparable offerings from Anthropic and OpenAI. Every new API account will begin with $20 in free credits, he said, after which Meta will charge $1.25 per million tokens for input and $4.25 per million tokens for output. The goal, Wang added, is “to really have attractive pricing that scales with immense consumption usage.”<\/p>\n
He also said Muse Spark 1.1 outperformed rival models on certain tasks involving interaction with third-party coding products and services. Meta Superintelligence Labs trained the model with a focus on coding because “you kind of have to build coding capabilities as part of that in service of overall agentic capabilities,” Wang said.<\/p>\n
The Muse Spark update is Meta’s second notable release in the Muse family this week. On Tuesday, Meta launched Muse Image, originally code-named Mango, an image-generation model aimed at attracting creators and advertisers.<\/p>\n
Tech industry interest in AI agents accelerated in the first half of 2026, partly because of the popularity of OpenClaw, which developers can use to manage AI models powering digital assistants. Wang said Meta trained Muse Spark 1.1 to work well with the most popular developer harnesses, calling it “the best approach for this model given our goal to maximize adoption.”<\/p>\n
Although Meta’s earlier AI efforts centered on releasing Llama family models to the open-source community, the company has shifted toward selling access to proprietary AI models. Wang said Meta remains committed to open source and that an in-development variant of Muse Spark is intended to be open-weight, though he declined to share a release timeline.<\/p>\n
Wang said Meta is training a more powerful AI model code-named Watermelon, though he did not share a release date. Muse Spark’s own code name was Avocado.<\/p>\n