Meta's Muse Wonder Full model has achieved a score of 56 on Artificial Analysis' Intelligence Index, tying with Anthropic's Claude Sonnet 5.5 and surpassing models from OpenAI and Alphabet. This positions Meta's latest AI models as frontier-class, a significant leap from its previous offerings.
TL;DR
- Meta's Muse Wonder Full model scores 56, tying with Anthropic's Claude Sonnet 5.5 in AI model rankings.
- Muse Wonder models represent a significant performance jump for Meta, with Muse Wonder Full outperforming models from OpenAI and Alphabet.
- Meta has increased the price of Muse Wonder and made Muse Wonder Full available only through subscriptions, indicating confidence in its new models.
What happened
Meta Platforms has announced that its latest Muse Wonder model family, particularly Muse Wonder Full, delivers state-of-the-art performance on personal and professional agentic tasks. According to Artificial Analysis, Muse Wonder Full scores 56, tying with Anthropic's Claude Sonnet 5.5 and ranking above OpenAI's GPT-6 Astra and Alphabet's Gemini 4 Argon, which both score 53. Muse Wonder scores slightly lower at 52. This marks a significant improvement from Meta's previous model, Muse Spark 1.3, which scored 48.
Meta has increased the price of Muse Wonder to $4.52 per task, 2.8 times higher than Muse Spark 1.3's cost of $1.60. Muse Wonder Full is only available through subscriptions to Muse Code, priced at $50 per month and $200 per month. This strategy aims to position Muse Wonder Full for agentic coding tasks and build a stable, subscription-based revenue stream.
Why it matters
This development signifies Meta's rapid progress in AI model development, moving from competitive to frontier-class in just over a year. It indicates Meta's growing capability to compete with industry leaders like OpenAI and Anthropic. For developers and startups, this could mean access to more powerful and cost-effective AI tools, while investors may see this as a positive sign for Meta's future AI monetization efforts.
However, the success of Muse Wonder and Muse Wonder Full will depend on their adoption and integration into Meta's products and services. The company's ability to translate these advancements into revenue growth will be a key focus for investors in the upcoming earnings report.
Key facts
- Muse Wonder Full scores 56 on Artificial Analysis' Intelligence Index, tying with Anthropic's Claude Sonnet 5.5.
- Muse Wonder scores 52, ranking above OpenAI's GPT-6 Astra and Alphabet's Gemini 4 Argon, both of which score 53.
- Muse Wonder's cost per task is $4.52, 2.8 times higher than Muse Spark 1.3's cost of $1.60.
- Muse Wonder Full is only available through subscriptions to Muse Code, priced at $50 per month and $200 per month.
- Meta's previous model, Muse Spark 1.3, scored 48 on Artificial Analysis' Intelligence Index.
- Meta began overhauling its AI efforts approximately 14 months ago, launching Meta Superintelligence Labs and hiring Alexandr Wang to lead the division.
- Meta's next earnings report is scheduled for October 21, 2026, after the market close.
Context
Meta's rapid advancement in AI model development comes as the tech industry continues to witness significant investments and innovations in AI. Companies like OpenAI, Anthropic, and Alphabet have been at the forefront of developing powerful AI models, driving competition and pushing the boundaries of what these models can achieve.
For Meta, this progress is crucial as it seeks to diversify its revenue streams beyond advertising. The company's focus on AI monetization is evident in its strategic moves, such as increasing model prices and introducing subscription-based services. However, the ultimate success of these efforts will depend on the adoption and integration of these models into Meta's products and services, as well as their impact on the company's financial performance.
