Meta Opens Muse Spark AI Model API to Developers
Meta Platforms has officially opened the API for its Muse Spark AI model to developers, alongside the release of an upgraded Muse Spark 1.1. This move signifies Meta's entry into the AI model commercialization race, charging for model calls.
Meta Platforms has officially made its Muse Spark AI model available to developers through its application programming interface (API), also releasing an upgraded version, Muse Spark 1.1. This strategic move positions Meta to compete directly in the commercialization of AI models, joining companies like Anthropic and OpenAI in charging for API access.
Muse Spark 1.1 is described by Meta as its most capable model to date for real-world coding tasks and AI agent capabilities. It is a key component of Meta's strategy to develop "personal superintelligence." The model can write and debug code, operate software and external tools, and interpret content from text, images, and video. It is also designed to handle complex, multi-step tasks with reduced human intervention.
Meta first introduced Muse Spark in April. The new API allows U.S. developers to test the model and integrate its functionalities into their applications. Developers receive $20 in free credits upon registration for initial testing, after which usage is priced based on call volume. The pricing is set at $1.25 per million input tokens and $4.25 per million output tokens, positioning it within the competitive range of other leading AI models.
In addition to the API, Muse Spark 1.1 is now live in the Meta AI application and its web version via a "Thinking" mode. The company plans to gradually replace its current Llama-series models, which power chatbots across Meta's platforms like WhatsApp, Instagram, and Facebook, with Muse Spark. This release follows Meta's recent expansion of generative AI features across its apps and the introduction of its first image generation model, Muse Image.