Meta Releases Muse Wonder AI Models

What is new

Meta has released two new AI models, Muse Wonder and Muse Wonder Full, which the company claims can compete with the industry’s best models. These models are part of Meta’s aggressive push to reestablish itself as a major force in AI. The new models are available for Muse Code subscribers and can also be accessed through the Meta Model API, with Wonder costing $4 per million input tokens and $20 per million output tokens.

The Muse Wonder and Muse Wonder Full models are larger and more capable than Meta’s previous model, Muse Spark, with higher benchmark scores and better results on complex tasks like coding. According to Meta’s testing, Wonder and Wonder Full achieve comparable benchmark scores to Fable 5.1 and GPT-6 Astra. The company has also emphasized the new models’ ability to retain and synthesize information given by the user, a property Meta calls “active memory”.

Why it matters

The release of Muse Wonder and Muse Wonder Full marks a significant step forward for Meta in the AI race. The company has poured billions into talent and infrastructure and has been steadily expanding its AI lineup, including with image model Muse Image and a heavy focus on bringing useful AI agents to the masses through Muse. The new models demonstrate Meta’s commitment to developing personal assistants that can perform tasks over a long period, with a particular emphasis on the ability to retain and synthesize information.

The ability of Muse Wonder and Muse Wonder Full to compete with leading models from OpenAI and Anthropic is a major milestone for Meta. The company’s testing has shown that the new models can achieve comparable benchmark scores to Fable 5.1 and GPT-6 Astra, and can even outperform these models in certain tasks. This demonstrates the significant progress Meta has made in developing its AI capabilities and positions the company as a major player in the AI industry.

What you can do with it

The Muse Wonder and Muse Wonder Full models can be used for a variety of tasks, including coding and other complex tasks. The models can be accessed through the Meta Model API, with Wonder costing $4 per million input tokens and $20 per million output tokens. Muse Code subscribers will also have access to the new models, and can use them to power the Muse agent.

The ability of the new models to retain and synthesize information given by the user makes them particularly useful for tasks that require a high level of contextual understanding. The models can be used to develop personal assistants that can perform tasks over a long period, and can even be used to power chatbots and other conversational AI systems.

Limits and open questions

While the release of Muse Wonder and Muse Wonder Full is a significant milestone for Meta, there are still several limits and open questions surrounding the new models. The company has not shared specifics or provided a timeline for when the new models will be used to power the Muse agent, and it is unclear how the models will stack up against Google’s Gemini 4 Argon, whose capabilities are still coming into focus.

Additionally, the cost of accessing the new models through the Meta Model API may be a barrier for some developers and users. The cost of $4 per million input tokens and $20 per million output tokens may be prohibitively expensive for some use cases, and may limit the adoption of the new models. However, the ability of the new models to compete with leading models from OpenAI and Anthropic makes them a significant development in the AI industry, and positions Meta as a major player in the field.