Meta Launches Muse Wonder, Claims AI Benchmark Leadership
What happened: Meta Superintelligence Labs released its new Muse Wonder and Muse Wonder Full AI models on October 12, 2026, touting state-of-the-art performance on internal benchmarks.
What happened: Meta Superintelligence Labs released its new Muse Wonder and Muse Wonder Full AI models on October 12, 2026, touting state-of-the-art performance on internal benchmarks. The models feature a 1 million token context window, active memory, and goal persistence. Meta’s own Personal Superintelligence Index ranks Wonder Full at 59.1, ahead of Anthropic’s Opus 5.5 (55.2) and OpenAI’s GPT-6 Astra (54.5), with competitive pricing at $4 per million input tokens. However, independent benchmarks from Artificial Analysis place Opus 5.5 ahead, with Wonder Full tied for second. The launch follows privacy concerns over Muse agent data collection.
Why it matters: Meta’s aggressive push into frontier AI models signals intensifying competition with Anthropic, OpenAI, and Google. While Meta claims leadership on its own metrics, external evaluations provide a more nuanced picture, with Opus 5.5 still leading on independent tests. The launch also raises ongoing privacy questions, as recent reporting highlighted data collection from users who opted out. The evolution of AI model performance and trust will continue to shape the competitive landscape in both consumer and enterprise AI.
Source: Decrypt