Muse vs Muse with Wonder: What's Different and Is the Upgrade Worth It?
Standard Muse runs on Muse Spark 1.3 and is free. Muse Wonder, Meta's new frontier model, powers Muse for $20–$100/month subscribers with longer memory, stronger reasoning, and better task persistence.
By Professor Muse6 min read
The short answer: Standard Muse runs on Muse Spark 1.3 and is free to use. Muse with Wonder runs on Meta’s new frontier model family, announced October 12, 2026, and requires a $20 or $100 per month subscription. You get longer-lasting memory across conversations, stronger benchmark performance, and an agent that keeps trying when tasks hit roadblocks.
Key takeaways
- Standard Muse runs on Muse Spark 1.3, Meta’s agentic foundation model with a ~1M-token context window. The app is free with a weekly usage allowance.
- Muse with Wonder upgrades the brain to Muse Wonder, Meta’s first frontier-class model. Access inside the Muse app requires the $20/month Power plan or the $100/month Maximum plan.
- The biggest practical differences are Active Memory (the model retains context about you across days, weeks, and months) and Active Focus (it persists through setbacks on long tasks).
- On independent benchmarks, Muse Wonder Full ranks in the top three on six of ten Artificial Analysis Intelligence Index evaluations, with particular strength in agentic knowledge work.
What powers standard Muse
Every free Muse download runs on Muse Spark 1.3, the first model family from Meta Superintelligence Labs. It is an agentic foundation model, built to act: browsing, calling, booking, and running errands across your connected apps, with a context window of about one million tokens.
The Muse app itself is free. Meta’s consumer tiers are usage-based: the free tier includes a weekly allowance (around 1M input tokens), the $20/month Power plan raises that substantially, and the $100/month Maximum plan is for heavy users. On the free tier you get the full agent (the same errand-running, phone-calling assistant) running on Spark 1.3.
What Muse Wonder changes
Muse Wonder, announced by Mark Zuckerberg on October 12, 2026, is Meta’s first frontier-class model family. It comes in two versions: Muse Wonder, the fast frontier model, and Muse Wonder Full, the full-strength version. Both are larger and more advanced than the Spark family, and both are available starting now.
Three things distinguish it in practice:
Active Memory. The model retains important context about you across conversations (days, weeks, months) and surfaces it when a later task needs it. Standard Muse works within its context window; Wonder remembers.
Active Focus. Meta trained Wonder to persist at difficult, long-horizon goals: trying different approaches when it gets stuck, asking for clarification, and checking with you before important actions. Meta AI chief Alexandr Wang pointed to tasks like booking dinner reservations and checking availability, where real-world chores hit snags.
Stronger safeguards. Meta says Wonder includes its strongest protections to date against cyber misuse. Those safeguards currently ship only with the paid models.
On benchmarks, the independent Artificial Analysis Intelligence Index places Muse Wonder Full (max) in the top three on six of its ten evaluations, noting particular strength in agentic knowledge work and scientific reasoning. Meta’s own charts compare it against Claude Fable 5.1, GPT-6 Astra, and Claude Opus 5.5.
Side-by-side comparison
| Standard Muse | Muse with Wonder | |
|---|---|---|
| Model | Muse Spark 1.3 | Muse Wonder (frontier-class) |
| Price | Free (weekly usage allowance) | $20/mo Power or $100/mo Maximum |
| Context window | ~1M tokens | ~1M tokens |
| Memory across conversations | Within-session context | Active Memory: retains context for weeks/months |
| Long-task persistence | Standard agentic behavior | Active Focus: tries alternate approaches before asking you |
| Benchmarks | Strong mid-tier performer | Top-3 on 6/10 Artificial Analysis evals |
| Safeguards | Standard | Strongest to date (paid only, for now) |
| Best for | Everyday errands, trying the agent | Power users, long-running projects, heavy agent workloads |
Pricing in detail
- Free: $0, the full Muse agent on Muse Spark 1.3, with a weekly usage allowance.
- Power: $20/month, unlocks Muse Wonder in the app, with much higher usage limits.
- Maximum: $100/month, same Wonder access, for the heaviest users.
- Muse Wonder Full (the strongest version) is not in the consumer app at all. It is exclusive to Muse Code subscribers on $50 or $200/month plans.
- Developers can also access Wonder through the Meta Model API at $4 per million input tokens and $20 per million output tokens.
Zuckerberg has said Meta will “work to bring this level of intelligence to everyone over time,” so the free tier may get Wonder eventually. For now, the frontier model is the paid tier’s headline feature.
Who should upgrade — and who shouldn’t
Upgrade if: you run long, multi-step projects through Muse and feel it losing the thread; you want the agent to remember context across weeks without re-explaining; you hit the free tier’s usage limits regularly; or you do the kind of agentic knowledge work where Wonder’s benchmark edge matters.
Stick with free if: your usage is everyday errands: bill calls, inbox triage, bookings, reminders. Muse Spark 1.3 handles those well, and the free agent hasn’t lost any capabilities. The subscription buys a smarter brain and more usage; the feature set stays the same.
How to get Wonder in your Muse
Subscribe to the Power ($20/month) or Maximum ($100/month) plan inside the Muse app. In the announcement, Zuckerberg said: “You can use Muse Wonder to power your Muse by subscribing today.” The rollout is happening now, so if you don’t see the option yet, check back. Meta is bringing it to subscribers in waves.
For the full launch details, see our coverage: Meta launches Muse Wonder, its first frontier model family.
Frequently asked questions
Do I have to pay to use Muse Wonder in the Muse app? Yes. Wonder powers Muse for subscribers on the $20/month Power plan or the $100/month Maximum plan. The free tier continues to run on Muse Spark 1.3.
What model does the free version of Muse use? Muse Spark 1.3, Meta’s agentic foundation model with roughly a million-token context window. The free app is the full agent (same errands, same features) with a weekly usage allowance.
What is Muse Wonder Full, and can I get it in the app? Wonder Full is the strongest version of the new model family, but it is exclusive to Muse Code subscribers ($50 or $200/month plans). Consumer Muse tops out at standard Muse Wonder.
What is Active Memory? Wonder’s long-term memory capability: the model retains important context about you across conversations over days, weeks, and months, and brings it back when a later task needs it.
What is Active Focus? Wonder’s persistence capability for long-horizon work. When a task hits a roadblock, the model tries different approaches on its own and checks with you before important actions.
Is Muse Wonder available in the app right now? Meta is rolling it out to subscribers now. Zuckerberg said you can power your Muse with Wonder by subscribing today. If the option isn’t visible in your app yet, it’s still propagating.
Will the free tier ever get Wonder? Possibly. Zuckerberg said Meta will “work to bring this level of intelligence to everyone over time.” No timeline has been announced.
