Meta Superintelligence Labs launched Muse Wonder and Muse Wonder Full, its first frontier models, aiming to compete with OpenAI and Anthropic on coding and memory tasks.
By Hermes-Vector AI Desk
Meta’s Superintelligence Lab released Muse Wonder and Muse Wonder Full on Monday, marking the company’s first attempt at a frontier AI model. The release follows a restructuring of Meta’s AI efforts led by Chief AI Officer Alexandr Wang, who joined the company after a $14 billion deal with Scale AI.
Muse Wonder is available to developers through the Meta Model API at $4 per million input tokens and $20 per million output tokens. According to TechCrunch, this pricing is roughly equivalent to OpenAI’s GPT-6 Sol and Anthropic’s Opus 5.5. The model features a one-million-token context window, with cached input costing $0.20 per million tokens, according to technology.org.
Muse Wonder Full, a larger and more capable variant, is not available via the API. It is reserved for subscribers to Meta’s Muse Code tool, specifically on the Power Usage and Maximum Usage plans, which cost $50 and $200 a month, respectively. Both models will power the Muse agent, an app that Meta says reached 5 million downloads in 22 days after its September launch.
Meta describes the new models as having "active memory," a feature designed to retain and synthesize user information over long periods. The company also introduced "active focus," which allows the models to continue working toward a goal when tasks become difficult.
According to Meta’s internal testing, Muse Wonder scores higher on the MemoryBench benchmark than Anthropic’s Fable, Opus, and OpenAI’s Astra. On the DeepSWE 1.1 coding test, Meta reports that Muse Wonder Full scored 79.0, compared to 74.2 for Opus 5.5 and 74.1 for GPT-6 Astra. Meta notes that these figures are self-reported and have not yet been independently verified.
Independent evaluator Artificial Analysis scored Muse Wonder Full at 56 on its Intelligence Index. This places it behind Anthropic’s Claude Opus 5.5, which scored 58, and close to GPT-6 Astra and Fable 5.1, which both scored 53. Meta’s blog post describes the models as "a major leap toward personal superintelligence."
Meta’s stock rose about 3.4% on Monday to close at $742.96, according to TIKR. The launch drew positive notes from several analysts. Wells Fargo said Meta has arrived at the "AI frontier," while KeyBanc stated the release "reinforces Meta has the right mix of talent, data, distribution, and compute capacity to compete in AI," according to CNBC.
CEO Mark Zuckerberg described the release as a turning point. "Now that we are kind of at the frontier, now everything that we do going forward will just kind of be pushing the frontier," Zuckerberg said on the Cheeky Pint podcast. Alexandr Wang emphasized that Meta is focusing on personal and professional agentic tasks, stating, "We do genuinely think that we've taken a differentiated focus to make the personal-agent use case really sing."
The launch represents Meta’s first major product from its massive capital expenditure on AI, which hit $30.12 billion in the second quarter of 2026. By introducing paid tiers for its most capable models, Meta is testing whether it can generate direct revenue from AI subscriptions and API usage, moving beyond its traditional advertising model. The success of Muse Wonder will be measured by its ability to compete with established leaders like OpenAI and Anthropic in both developer adoption and financial returns.
Sources
This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.
The day's checked news, by email
One short email a day with the top stories. Free, and you can unsubscribe in one click.
Or follow for the latest news:
The U.S. Air Force has awarded WindBorne Systems a contract worth up to $49 million for atmospheric data from its autonomous balloon fleet. The agreement allows the Air Force to order data as needed, with $1.7 million in orders placed so far.
Representatives from OpenAI and Anthropic told a British parliamentary committee that voluntary AI safety pledges are insufficient, while Google DeepMind and Meta declined to support new statutory frameworks.
