Meta Unveils Its First Advanced AI Model

Meta's Superintelligence Labs unit has released a pair of new AI models, Muse Wonder and Muse Wonderful, marking the company's first effort to deliver a frontier model.

Meta's "Super Intelligence Lab" unit has released a pair of new artificial intelligence models named "Muse Wonder" and "Muse Wonderful", which represent the company's first efforts to provide a leading model.

According to Ambassador Arjan, quoted by TechCrunch, while the "Muse Spark" was supposed to be a lightweight model that could work with little delay, the Wonder models are larger and according to information more powerful and also have a better benchmark score in performing complex activities such as coding.

Based on meta tests, Wonder and Wonderful models have achieved a benchmark score equivalent to Astra's Fable 5.1 and Chat GPT 6 models. Together, they represent the next step on the scaling ladder and a giant leap toward personal superintelligence, Meta's blog post about these models says.

The said scale is reflected in the pricing. The Wonder model is accessible through the "Metamodel API" and its price is $4 per million input tokens and $20 per million output tokens. In this way, its cost is almost similar to the pricing of Chat GPT 6 Soul and Opus 5.5.

Interestingly, the Wonderful model is only available to subscribers of the "Muse Code" plan and is not offered through API. Both models are used to activate Muse Agent.

In line with its interest in developing personal assistants that can perform activities over long periods of time, the company's goal is to emphasize the model's ability to store and synthesize information provided by the user. Meta calls this feature "active memory".