Meta has launched the next version of its in-house AI model for coding, making it available to developers through a public API preview.
The company says Muse Spark 1.1 represents a significant improvement over the first generation, which debuted in April. The updated model is designed to handle more advanced coding tasks, including detecting and fixing complex bugs. It also supports end to end agentic workflows across multiple applications, including systems that use multiple AI agents working together. The model now has native multimodal perception, allowing it to process images, videos, and documents.
Meta Releases Muse Spark 1.1 for Developers
The launch follows a busy period for Meta's AI efforts. Earlier this week the company released Muse Image, an image generation model that sparked controversy over its ability to pull content from Instagram users' posts. Both models are part of Meta's broader push to justify the billions it has spent catching up with rivals like OpenAI, Google, and Anthropic. Meta underwent a major restructuring last year and made several high profile hires in the AI space to accelerate its progress.
Muse Spark 1.1 is now available in Thinking mode through the Meta AI app and the Meta AI website. For the first time, developers can access the model via a new Meta Model API, which entered public preview today for developers in the United States. Each new Meta Model API account comes with $20 worth of free credits to get started.
The original Muse Spark model was initially only available directly through Meta AI. It later powered the chatbots inside Instagram and WhatsApp, as well as the latest version of Meta's smart glasses.
Coding and Multimodal Improvements
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Meta describes Muse Spark 1.1 as a "step-change" from its predecessor. The improvements were based on feedback from developers who had worked with the first version. The model's coding capabilities have been expanded to include more sophisticated tasks. Its end to end agentic workflow support means it can handle complex, multi step instructions across different apps and systems, including scenarios where multiple AI agents collaborate.
The addition of native multimodal perception allows the model to understand and generate responses based on images, videos, and documents, not just text. This makes it more versatile for developers building applications that require visual understanding alongside code generation.
Availability and Pricing
The model is also accessible to consumers through the Meta AI app and website, where it can be used in Thinking mode. This mode is designed for more complex reasoning tasks. Meta has not disclosed pricing for the API beyond the initial free credits, or whether it will offer tiered plans.
Muse Spark 1.1 arrives as Meta continues to invest heavily in AI infrastructure and talent. The company's spending in this area has drawn scrutiny from investors, but Meta has signaled it plans to maintain its pace. The new model represents a direct challenge to coding focused AI offerings from competitors like OpenAI's Codex and Google's Gemini for coding.
Related on Neura Market
- AI Model Directory, Explore and compare the latest AI models for coding, image generation, and more.
- Coding Automation Marketplace, Find tools and APIs to integrate AI into your development workflow.
- Developer Tools Platform, Browse platforms that offer APIs and SDKs for building AI powered applications.