Meta Releases AI Models for Document and Code Processing
The new Muse Wonder and Muse Wonder Full models can analyze insurance policies, lease agreements, and financial reports
Meta has introduced two new artificial intelligence models—Muse Wonder and Muse Wonder Full. They are designed to tackle professional tasks in law, finance, and medicine, as well as for writing and reviewing application and game code.
The models feature an 'active memory' function, which retains context between conversations and retrieves relevant details for the current task. Another function, 'active focus,' allows the AI to continue working even if the task becomes more complex, try different approaches, correct errors, and clarify details before performing critical actions.
The company emphasizes that the models are trained to protect users' personal data and refuse to execute dangerous requests. They can also assist developers and security specialists in identifying vulnerabilities in code and strengthening security systems. All models undergo evaluation according to Meta's Superintelligence Scaling Framework standards.
Muse Wonder is already available via Meta's API and the Muse Code service. Subscribers will gain access to the new capabilities shortly. The cost of using the model is 4 dollars per million input tokens and 20 dollars per million output tokens. Muse Wonder Full is available to users on the 'extended use' (50 dollars per month) and 'maximum use' (200 dollars per month) plans.
Meta CEO Mark Zuckerberg, in an interview with the Financial Times, noted that developing the models required significant computational power and the scientific work of the company's superintelligence team. This year, Meta plans to spend up to 145 billion dollars on AI technology development.
