Meta Releases Muse Wonder, Its First Frontier AI Model
Disclosure: this post contains affiliate links. If you buy through them we may earn a commission at no extra cost to you. Learn more.
TL;DR: Meta's Superintelligence Lab released Muse Wonder and Muse Wonder Full on October 12, 2026, marking the company's first frontier model release. The models achieve benchmark scores comparable to Fable 5.1 and GPT-6 Astra, with particular emphasis on 'active memory' capabilities for long-term information retention.
Meta Releases Muse Wonder, Its First Frontier AI Model
Meta's Superintelligence Lab unveiled Muse Wonder and Muse Wonder Full on Monday, October 12, 2026, representing the company's first attempt at a frontier model. This release marks a significant shift for Meta, which has previously focused on lightweight models rather than competing directly in the state-of-the-art model space.
What's New
The Muse Wonder models are substantially larger and more capable than Meta's previous Muse Spark offering, which prioritized low latency over raw performance. According to Meta's model card testing, Wonder and Wonder Full achieve benchmark scores comparable to Fable 5.1 and GPT-6 Astra, with particularly strong results on complex tasks like coding.
A standout feature of the new models is what Meta calls "active memory"—the ability to retain and synthesize information given by users over extended periods. "We put a tremendous amount of effort into making sure your model can remember what's important to you over days, weeks, months, and beyond," said research scientist Kavya Kopparapu in a video accompanying the launch.
Meta highlighted the models' performance on MemoryBench, a benchmark that challenges continual learning and memory construction frameworks. According to the company's testing, Muse Wonder now scores higher on MemoryBench than Fable, Opus, or Astra.
Both models will power the Muse agent, aligning with Meta's broader interest in developing personal assistants capable of performing tasks over long periods. Meta's blog post about the models states they "mark the next step on our scaling ladder and a major leap toward personal superintelligence."
The models were trained at Meta's Prometheus data center in central Ohio, which came online earlier in 2026.
Pricing and Availability
Muse Wonder is accessible through the Meta Model API with pricing set at $4 per million input tokens and $20 per million output tokens, making it roughly equivalent in cost to GPT-6 Sol or Opus 5.5.
Muse Wonder Full will only be available to Muse Code subscribers and will not be served through the API. Both models became available on October 12, 2026.
How to Try It
Developers can access Muse Wonder through the Meta Model API. The model is also integrated into the Muse agent for users who want to leverage its capabilities through Meta's assistant interface.
For Muse Wonder Full, users will need a Muse Code subscription, as it is not available through the standard API.
Who It's For
Muse Wonder targets developers and organizations needing advanced AI capabilities for complex tasks, particularly coding and applications requiring long-term memory. The model's emphasis on active memory makes it especially suitable for personal assistant applications and use cases where context retention over extended periods is crucial.
Muse Code subscribers seeking the most capable version of the model can access Wonder Full, though pricing for the subscription service was not disclosed.
Alternatives
Meta positions Muse Wonder as competitive with other frontier models including Fable 5.1, GPT-6 Astra, GPT-6 Sol, and Opus 5.5. Organizations evaluating frontier models for development work may also consider tools like Replit for AI-powered coding environments.
This release represents Meta's first frontier model launch since Scale AI founder Alexandr Wang restructured Meta's AI efforts in August 2025. Since that restructuring, Meta had released the lightweight Muse Spark model and Muse agent but had avoided the kind of state-of-the-art model releases that competitors like OpenAI and Anthropic are known for.
Tools mentioned
FAQ
What is Muse Wonder?
Muse Wonder is Meta's first frontier AI model, released on October 12, 2026. It's a large, state-of-the-art model that achieves benchmark scores comparable to Fable 5.1 and GPT-6 Astra, with particular strengths in coding and long-term memory retention.
How much does Muse Wonder cost?
Muse Wonder costs $4 per million input tokens and $20 per million output tokens through the Meta Model API, making it roughly equivalent in cost to GPT-6 Sol or Opus 5.5. Muse Wonder Full is only available to Muse Code subscribers and not through the API.
What is 'active memory' in Muse Wonder?
Active memory is Meta's term for the models' ability to retain and synthesize information given by users over days, weeks, months, and beyond. Muse Wonder scores higher on the MemoryBench benchmark than competing models like Fable, Opus, or Astra.
What's the difference between Muse Wonder and Muse Wonder Full?
While specific performance differences weren't disclosed, Muse Wonder is available through the Meta Model API at standard pricing, while Muse Wonder Full is only available to Muse Code subscribers and cannot be accessed through the API.