Meta has released two new AI models, Muse Wonder and Muse Wonder Full, designed for professional and personal use. On the Artificial Analysis Intelligence Index v4.3, the Muse Wonder Full model achieved a score of 56, while the standard Muse Wonder reached 52. This marks an improvement over the previous Muse Spark 1.3 model, which scored 48.
The Muse Wonder Full model currently holds the second position on the index, trailing only the Claude Opus 5.5 model from Anthropic. In specific categories, including coding, legal, and automation tests, the new models reportedly outperform GPT-6 Astra and Claude Fable 5.1. Muse Wonder Full placed in the top three across six of the ten index evaluations.
To improve reliability in agentic tasks, Meta introduced two new features: Active Memory and Active Focus. These additions are intended to enhance the dependability of the models during complex operations. The company is positioning this new lineup as a direct competitor to existing leaders in the AI industry, such as OpenAI and Anthropic.
The pricing for the Muse Wonder model is set at $4 per million input tokens and $20 per million output tokens. These releases follow a strategic restructuring of Meta’s AI development efforts that began in 2025. The company aims to leverage these performance gains to challenge the current market landscape.
