Meta releases its first frontier model, Muse Wonder
techcrunch ·
Meta’s Superintelligence Lab has unveiled its first frontier models, named Muse Wonder and Muse Wonder Full, marking a significant shift in the company's AI strategy. These new large-scale models are designed to outperform previous lightweight options like Muse Spark, offering higher benchmark scores and improved capabilities in complex tasks such as coding. Meta claims that these models achieve performance comparable to industry leaders Fable 5.1 and GPT-6 Astra. A key feature highlighted is 'active memory,' which allows the models to retain and synthesize user information over extended periods, a capability tested via the MemoryBench benchmark. The models are priced at $4.00 per million input tokens and $20.00 per million output tokens, positioning them competitively against other high-end offerings. While Muse Wonder is available through the Meta Model API, the more powerful Muse Wonder Full is restricted exclusively to Muse Code subscribers. This release follows a restructuring of Meta's AI efforts last year and represents the company's entry into the state-of-the-art model market previously dominated by competitors like OpenAI and Anthropic.
techcrunch