Meta Platforms has launched two new artificial intelligence models, Muse Wonder and Muse Wonder Full, as it steps up efforts to compete with leading AI developers OpenAI and Anthropic in the race to build more capable systems.

Announced on Monday, October 12, the models are designed to handle complex tasks, including software development, and improve how AI assistants remember users’ preferences and information shared across conversations. Meta said the models deliver state-of-the-art performance on benchmarks for coding, memory and tasks involving AI agents.

“Together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence,” Meta said in its announcement.

The launch marks another step in Meta’s push to develop more powerful AI systems, following the introduction of Muse Spark, a smaller model designed to deliver responses quickly. The company said the new models were trained at its Prometheus data centre in Ohio.

Muse Wonder is available through Meta’s Model API and Muse Code, the company’s coding tool. Muse Wonder Full, the more advanced version, is currently restricted to Muse Code subscribers, with no API access announced.

Meta said Muse Wonder will also power experiences in its Muse assistant, initially for subscribers, as well as tasks requiring greater intelligence, such as generating digital artefacts.

A key feature of the new models is what Meta calls “Active Memory”, which allows an AI system to retain important information about a user across conversations and retrieve it when relevant to a task.

For instance, an AI assistant could remember a user’s preferences, plans and earlier instructions, allowing it to offer more tailored responses without requiring the person to repeat the same details each time.

Source link