On October 12, Meta introduced two new AI models: Muse Wonder and Muse Wonder Full. The company calls them its most capable models for agentic tasks, meaning work where the AI doesn't just answer a question but takes a task all the way to a result: reviewing documents, writing and fixing code, analyzing data.
What the new models can do
There are two headline features.
- Active Memory. The model remembers important context about the user across conversations and pulls in the relevant details for a new task. No more re-explaining the project, what you agreed on and what's already done every time you come back.
- Active Focus. The model stays on the goal and doesn't give up after the first error: it tries another approach, works around blockers, asks for clarification when needed and checks in before important actions. For example, it writes code, runs a check, sees an error, fixes it and checks again.
According to VentureBeat, the models accept text, images, audio and video as input and have a context window of about 1 million tokens. They respond in text.
Availability and pricing
- Muse Wonder is live in the Meta Model API and in Muse Code. API pricing is $4 per 1M input tokens and $20 per 1M output tokens, with cached input at $0.20 per million.
- Muse Wonder Full isn't in the API. It's only available to Muse Code subscribers. According to VentureBeat, that means the $50 and $200 monthly plans.
- The Muse app. Muse Wonder will come to it later, with subscribers getting access first. Meta hasn't given dates.
On the independent Artificial Analysis index, Wonder Full scored 56 and the standard Wonder scored 52. Meta separately highlights results on long-running coding and legal benchmarks, but some of those evaluations were run by Meta itself, so treat them with caution.
The detail few people mention: your data
Muse Code subscriptions have Contributor mode turned on by default. In this mode, Meta may use your prompts, code, files and model outputs to train future models. You can turn it off, but only for new submissions: whatever you've already sent stays with Meta. The API runs in Standard mode, where data isn't used for training.
This matters for media buyers. Don't upload your funnels, creatives, campaign reports, client data or credentials to Muse Code with default settings. Turn off Contributor mode first or work through the API.
What it means for affiliate marketing
The launch doesn't affect Facebook and Instagram ads yet: Meta hasn't announced any Ads Manager integration or changes to ad delivery algorithms. But as a work tool, Muse Wonder can already be useful:
- Analytics. Combine exports from your tracker and ad accounts, compare funnels and prepare a summary for the team.
- Internal tools. Draft a reporting script, a scraper or a simple dashboard. This kind of multi-step work is exactly where the model is supposed to shine.
- Documents. Review affiliate program terms, an agency contract or platform rules and flag the questionable clauses.
How well the models hold up in real work will become clear after a couple of weeks of testing. If you already use AI agents, Muse Wonder is worth trying through the API: it's priced no higher than other top models, and memory across sessions saves a lot of time on long projects.