Meta Launches Muse Spark 1.1 to Compete in AI Coding Market
Meta has introduced Muse Spark 1.1, an update to its artificial intelligence model aimed at coding and agentic workloads, as the company looks to compete with OpenAI and Anthropic in a fast-growing segment of the AI market. AI chief Alexandr Wang described the release as Meta’s strongest model yet for agentic and coding work.
What is Muse Spark 1.1 and how much does it cost?
The original Muse Spark model, released in April, was limited to select partners through a private API preview. Muse Spark 1.1 is now accessible through a public developer portal, where users can sign up, view integration instructions, and join a waitlist. Meta said some early partners already have API access, and new users will be onboarded over time.
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 said the model is served on infrastructure Meta has built internally.
Meta priced the new model aggressively relative to competitors, according to Wang. Every new API account receives $20 in free credits, followed by $1.25 per million tokens for input and $4.25 per million tokens for output.
Why is Muse Spark 1.1 focused on coding?
Wang said Muse Spark 1.1 was trained by Meta Superintelligence Labs (MSL) to perform well on coding-related tasks, because coding capabilities are foundational to AI agents that can autonomously complete multi-step work. “You kind of have to build coding capabilities as part of that in service of overall agentic capabilities,” he said.
Wang claimed the model outperformed rival systems on certain tasks involving interaction with third-party coding tools and services. He added that Muse Spark 1.1 was designed to work with the most popular developer harnesses in use today, a choice driven by the goal to maximize adoption.
Industry interest in AI agents accelerated in the first half of 2026, partly due to the sudden popularity of OpenClaw, a tool developers use to manage AI models powering digital assistants.
Why is Meta moving from open source to proprietary access?
Meta’s earlier AI strategy centered on releasing its Llama family of models to the open-source community. The Muse Spark rollout marks a different approach: selling access to proprietary models. Wang said Meta remains committed to open source and that MSL has a variant of Muse Spark in development that it intends to release as open-weight, though he declined to share a timeline.
What else is in Meta’s Muse roadmap?
The Muse Spark update is Meta’s second major Muse release of the week. On Tuesday, the company launched Muse Image, previously code-named Mango, a model for image generation aimed at creators and advertisers.
Wang also confirmed that Meta is training a more powerful model, code-named Watermelon, though no release date has been disclosed. Muse Spark itself was internally code-named Avocado.
Beyond coding, Wang said he has been personally testing Muse Spark 1.1 for health-related tasks such as searching the web, reading academic papers, and accessing personal health data, calling it a use case that encapsulates the needs of agentic systems.
What business pressures is Meta responding to?
Meta faces ongoing pressure from Wall Street to demonstrate returns on its substantial AI infrastructure investments. While the company spends at a rate comparable to hyperscaler peers, it does not yet operate a cloud infrastructure business, though it has indicated plans to launch one. Competitors such as OpenAI, Anthropic, and Google have maintained leads in shipping widely adopted models and applications.
FAQ
What is Muse Spark 1.1?
Muse Spark 1.1 is Meta’s updated AI model for coding and agentic workloads, released by Meta Superintelligence Labs and described by AI chief Alexandr Wang as Meta’s strongest model yet for those tasks.
How can developers access Muse Spark 1.1?
Developers can sign up through Meta’s public developer portal to view integration instructions and join a waitlist, with new users onboarded over time. Access is currently limited to Meta’s own properties and is not distributed via third-party marketplaces such as OpenRouter.
How much does the Muse Spark 1.1 API cost?
Every new API account receives $20 in free credits, with input priced at $1.25 per million tokens and output at $4.25 per million tokens, which Wang described as aggressive relative to competitors.