Meta's Superintelligence Lab released two models on October 12, Muse Wonder and Muse Wonder Full, which TechCrunch's Russell Brandom describes as Meta's first attempt at a frontier model. Both will power the Muse agent. Muse Spark was a lightweight model built for low latency, and the Wonder models are larger and, per Meta, more capable, with higher benchmark scores and better results on complex tasks such as coding.
Availability and pricing
Muse Wonder is accessed through the Meta Model API at $4 per million input tokens and $20 per million output tokens, which the article says is roughly equivalent in cost to GPT-6 Sol or Opus 5.5. Muse Wonder Full is available only to Muse Code subscribers and is not served through the API.
Meta's benchmark claims
According to Meta's model card, Wonder and Wonder Full achieve benchmark scores comparable to Fable 5.1 and GPT-6 Astra, and Muse Wonder scores higher on MemoryBench than Fable, Opus or Astra. TechCrunch presents all of these as Meta's own, not independently verified, and names no independent evaluations. The article includes no safety remarks or safety testing details.
The memory emphasis
Meta highlights active memory, the models' ability to retain and synthesize information users provide, aimed at long-running personal assistant tasks. Research scientist Kavya Kopparapu said in a launch video that the goal is making sure the model can remember what is important to you. Meta's blog post said: “Together, they mark the next step on our scaling ladder and a major leap toward personal superintelligence.”
Context
This is the first frontier release since Alexandr Wang restructured Meta's AI efforts in August 2025. Since then Meta has released Muse Spark and the Muse agent but not the kind of state-of-the-art models associated with OpenAI and Anthropic. Meta says the models were trained at its Prometheus data center in central Ohio, which came online earlier in 2026.
Source: TechCrunch. This is an AI Business Watch summary of the original report.