The release of developer tools for Meta's new artificial intelligence (AI) model appears to be delayed.
According to the Wall Street Journal (WSJ) on the 4th (local time), the outlet, citing multiple sources, said Meta has repeatedly postponed plans to release developer tools for the AI model "Muse Spark," which it released in early April, and has yet to set a specific timeline.
For developers to use AI models that are not open-source, unlike "chatbot" formats for general users, they must use API (application programming interface) tools that allow communication with other programs. AI model developers prepare such API tools and generate revenue by measuring how many tokens developers use and charging accordingly.
Alexander Wang, Meta's chief AI officer (CAIO), also signaled the release of an API via X (formerly Twitter) right after unveiling Muse Spark. But nearly two months after the model's release, Meta still has not rolled out an API version of the model. WSJ reported that the delay in releasing the Muse Spark API may be disrupting Meta's plans to pursue monetization through its AI model.
Before developing Muse Spark, Meta had mainly focused on open-source models such as Llama, so there was little need to release API tools. Because users could download open-source models and install them on their own servers, there was no reason to route through Meta's servers via an API.
Because Muse Spark is Meta's first closed model, entering the developer ecosystem via an API is crucial. In particular, Meta, a latecomer in AI models, released that Muse Spark showed benchmark performance comparable to the latest models from OpenAI and Anthropic, so an API must be released to verify whether such performance can be reproduced in real-world use.
Meta said, "We know users want an API," and noted, "We are testing the API with partners and plan to release it this month."