MENLO PARK: Meta launched the Muse Wonder AI models on Monday, introducing two systems for digital agents and software development as it seeks to compete with OpenAI and Anthropic.
The company described Muse Wonder and Muse Wonder Full as its most capable artificial intelligence models to date. Meta offers Muse Wonder through the Meta Model API and Muse Code, while the more powerful Muse Wonder Full requires a Muse Code subscription.
Meta priced Muse Wonder at $4 per million input tokens and $20 per million output tokens. Both models support a 1 million-token context window.
Meanwhile, Muse Wonder Full is available through Muse Code’s Power Usage and Maximum Usage subscriptions, costing $50 and $200 per month respectively.
Read: Meta Muse Targets Subscription Spending With AI Agent
Meta plans to integrate Muse Wonder into its Muse personal AI agent, initially for subscribers. However, it has not announced a date for wider availability.
The company said its models can manage complex tasks involving software development, financial documents and legal information.
It introduced Active Memory to retain relevant details across conversations and Active Focus to help agents complete lengthy tasks.
Read: OpenAI Releases ChatGPT 5.5 With Stronger Agentic AI Capabilities
According to Meta, these features let its systems adjust their approach, recover from errors and seek clarification before taking important actions. However, the company’s performance comparisons with competing AI models rely on its own evaluations.
Chief Executive Mark Zuckerberg has invested heavily in AI infrastructure and recruited researchers to strengthen Meta’s position against rivals.
Read: AI Agents Get Cute Mascots as OpenAI Launches Dots
The company began restructuring its AI operations 14 months ago. Meta shares gained about 4% in Monday afternoon trading as investors digested the announcement. Meta said it would publish additional safety evaluations of the models, but it did not provide a release date.
