Punjab Newsline | New Delhi
Artificial intelligence (AI) is rapidly evolving beyond answering questions, with Meta introducing two new AI models, Muse Wonder and Muse Wonder Full, designed to handle complex tasks involving multiple steps. According to the company, the models can work on tasks such as understanding legal documents, conducting financial analysis and developing software, reducing the need for users to provide separate instructions at every stage.
Meta describes the models as its most capable AI systems to date, developed to assist users with complex tasks in both their personal and professional lives. Their potential applications include legal work, financial analysis, software development and business workflow automation.
Unlike conventional AI tools, which often require users to provide follow-up instructions after each response, the new models are designed to handle longer, multi-step assignments. For instance, they could help users understand lease agreements, process insurance claims or build applications by working through different stages of a task.
Meta CEO Mark Zuckerberg described the launch as a significant milestone in the company's AI efforts. In a post on social media platform X, he said Muse Wonder represents a major step towards bringing personal superintelligence to everyone. The company's broader goal is to develop AI assistants that can provide more capable support across personal and professional activities.
How Did the Models Perform in Benchmark Tests?
Meta also shared results from three benchmarks to demonstrate the capabilities of its new models. Benchmarks are standardised tests used to evaluate AI systems on specific tasks.
In the Harvey LAB-AA legal agent benchmark, Muse Wonder Full scored 13.1, while Muse Wonder recorded 10.0. Other models, including GPT-6 Astra, scored 8.6, according to figures shared by Meta.
In the DeepSWE 1.1 long-horizon coding benchmark, Muse Wonder Full achieved a score of 79.0, narrowly outperforming Muse Wonder, which scored 78.8.
Meanwhile, in the Automation Bench-AA business workflow benchmark, Muse Wonder Full scored 45.4, compared with 39.1 for Muse Wonder. The results indicate that Muse Wonder Full performed better than its counterpart across all three benchmarks cited by the company.
Meta said the models are intended to make AI more capable of managing end-to-end tasks rather than simply responding to individual prompts. However, the benchmark figures were shared by Meta, and independent evaluations would help establish how the models perform in real-world applications.
