Meta launched developer access to its AI model Muse Spark and also released an upgraded version called Muse Spark 1.1 on Thursday, according to Meta. This is the first time outside developers can use the model through Meta's public API. With this launch, Meta is directly competing with AI companies OpenAI and Anthropic by charging developers to use its AI models.
Meta described Muse Spark 1.1 as its strongest AI model for coding and agentic tasks. Agentic AI means the model can complete tasks on its own with much less human help. Meta said Muse Spark 1.1 can write new code, fix coding errors, understand software, and solve programming problems. The model can understand text, images, and videos together because it is a multimodal AI model. The launch shows Meta is pushing harder to compete with OpenAI and Anthropic in the fast-growing AI market.
Muse Spark 1.1 features
Meta said the AI can also use software and external tools to complete tasks automatically. The company said Muse Spark 1.1 can carry out difficult multi-step tasks without needing constant human instructions. Meta launched the first Muse Spark reasoning model in April after creating its Superintelligence team last year to compete with leading AI companies.
Meta model API launch
The company also launched a public preview of the Meta Model API, allowing developers to build apps using Muse Spark. An API works like a digital bridge that lets developers connect Meta's AI model to their own software, according to Reuters. Developers in the United States can now test Muse Spark through the Meta Model API.
{{/usCountry}}The company also launched a public preview of the Meta Model API, allowing developers to build apps using Muse Spark. An API works like a digital bridge that lets developers connect Meta's AI model to their own software, according to Reuters. Developers in the United States can now test Muse Spark through the Meta Model API.
{{/usCountry}}Also read: OpenAI launches GPT-Live voice models that can speak and listen at the same time
They can test prompts, compare AI responses, and build new applications before launching them. Every new developer who signs up receives $20 in free credits to test the model before paying for usage. After the free credits end, Meta charges $1.25 for every one million input tokens and $4.25 for every one million output tokens.This means the AI can remember and process much longer documents and conversations at one time. Meta's pricing is higher than OpenAI's GPT-5 mini and Anthropic's Claude Haiku 4.5, but cheaper than Anthropic's Claude Sonnet 4.6, according to Reuters.
AI apps and coding
Muse Spark 1.1 is now available in "Thinking" mode on the Meta AI app and Meta's website. Meta also plans to replace the current Llama AI models used in WhatsApp, Instagram, Facebook, and Meta smart glasses with Muse Spark. Earlier this week, Meta also launched Muse Image, its first image-generation model from Meta Superintelligence Labs.
Meta said Muse Spark 1.1 comes with a huge one-million-token context window, allowing it to understand and remember much larger amounts of information in one conversation. The company said the model is designed to work inside multi-agent systems where multiple AI agents can work together on the same task. Meta said Muse Spark 1.1 can act as both the main AI agent and a supporting AI agent at the same time. This allows the model to divide a large task into smaller jobs and complete them faster by running them in parallel.
Better coding performance
Meta said the upgraded model performs much better on very large enterprise coding projects than its earlier version. According to Meta, the AI can automatically find software bugs, fix them, add new features to old software, and help companies move large amounts of code to newer systems. Meta also introduced new computer-use features in Muse Spark 1.1.
The AI can automate work across different desktop applications by creating its own automation scripts. If writing scripts is not possible, the AI can directly interact with software interfaces to complete tasks. Several companies have already started using the API during its early rollout.
Safety and early partners
The first major partners include Replit, Cline, and Box. Replit CEO Amjad Masad said the combination of a one-million-token context window, multimodal abilities, and advanced coding features could change how developers build software at scale, according to Investing.com.
Cline CEO Saoud Rizwan said the model's flexible tool use and competitive pricing make it attractive for developers handling large coding workloads. Box Vice President of AI Products Yashodha Bhavnani said Muse Spark 1.1 performed on the same level as some of the industry's leading AI models during Box's internal testing, according to Investing.com.
Meta said it also carried out extensive safety testing before releasing the model. The safety checks looked for risks involving chemical and biological misuse, cybersecurity threats, and situations where AI systems could lose control. Meta said Muse Spark 1.1 stayed within its safety limits and that a detailed safety report will be released soon.
The launch of Muse Spark 1.1 and Muse Image shows Meta is expanding its AI products for both consumers and businesses while increasing competition with OpenAI and Anthropic.