Models
What Is Muse Wonder? Meta’s First Frontier Model, Explained
Muse Wonder is Meta’s first frontier AI model, launched October 12, 2026. Muse Wonder vs Muse Wonder Full, price, benchmarks, and how it compares to Opus 5.5.
Muse Wonder is Meta’s most powerful AI model so far. Meta launched it on October 12, 2026, together with a bigger sibling called Muse Wonder Full. Meta calls them its first "frontier" models, meaning they are built to compete with the very best from OpenAI, Anthropic, and Google.
The names are confusing, because Meta now has Muse Spark, Muse Spark Open, Muse Wonder, and Muse Wonder Full. So here is what Muse Wonder actually is, what the "Full" version adds, what it costs, and how it stacks up.
What is Muse Wonder?
Muse Wonder (API name muse-wonder-1.0) is a large language model made by Meta’s AI lab, Meta Superintelligence Labs, which is led by Alexandr Wang. It is the lab’s first frontier release since Wang reorganized Meta’s AI work in August 2025.
Meta did not pitch it as a chat model. It pitched it as an engine for agents: AI that writes code, uses tools, and keeps working through a long job without a person nudging it at every step. Meta says it was trained at its new Prometheus data center in central Ohio.
Two features get the most attention:
- Active Memory. The model carries what matters from one conversation to the next: your goals, rules you set, and decisions made earlier.
- Active Focus. On hard tasks, it tries other approaches when one fails, asks for clarification, and checks in before an important action.
Both models can read up to 1 million tokens at once and accept images as well as text. Neither one is open weight: you cannot download them. If you want a Meta model you can run yourself, that is Muse Spark Open.
Muse Wonder vs Muse Wonder Full
This is the part that trips people up. They are two separate models, sold in two different ways.
| Muse Wonder | Muse Wonder Full | |
|---|---|---|
| Released | October 12, 2026 | October 12, 2026 |
| Strength | Strong, near the top tier | Meta’s strongest model |
| Context window | 1 million tokens | 1 million tokens |
| Meta Model API | Yes, $4 in / $20 out per million tokens | No |
| Muse Code (coding app) | Yes | Yes, on Power Usage ($50/month) and Maximum Usage ($200/month) plans |
| Open weights | No | No |
In short: Muse Wonder is the one businesses can build on. Muse Wonder Full is the one you can only reach by paying for Muse Code, Meta’s coding agent product. VentureBeat summed it up in its headline: only the weaker one is available through the API.
Muse Wonder price
For developers, Meta lists these API prices for Muse Wonder:
- $4.00 per million input tokens
- $0.20 per million cached input tokens
- $20.00 per million output tokens
- Web search: $2.50 per 1,000 searches, billed separately
That is the same token price as Claude Opus 5.5, as TechCrunch and VentureBeat both pointed out. A token is a small piece of a word, so one million tokens is roughly 750,000 words.
Is Muse Wonder free? Not right now. Meta has not announced a free way to use it. The API is pay as you go, and Muse Wonder Full needs a Muse Code plan. Analyst firm Futurum reports that both models will also power Meta’s Muse personal agent app, starting with subscribers on its $20 and $100 monthly plans. Meta has not shared a date for that.
Muse Wonder benchmarks: how good is it?
Meta’s own model card says Wonder and Wonder Full perform about as well as Claude Fable 5.1 and GPT-6 Astra, and that Wonder Full leads on two long coding tests, DeepSWE and SWE-Marathon. Those are Meta’s numbers, and TechCrunch noted it had not checked them.
The independent Artificial Analysis Intelligence Index gives a cooler picture. These scores were reported by both VentureBeat and Futurum:
| Model | Artificial Analysis Index | Can developers use it via API? |
|---|---|---|
| Claude Opus 5.5 | 58 | Yes |
| Claude Sonnet 5.5 | 56 | Yes |
| Muse Wonder Full | 56 | No |
| Claude Fable 5.1 | 53 | Yes |
| GPT-6 Astra | 53 | Yes |
| Muse Wonder | 52 | Yes |
So the honest verdict: Muse Wonder Full is a real top-tier model, and memory is its standout skill. But the version you can actually build on, plain Muse Wonder, sits a step below Opus 5.5 at the same price. For Meta, it is a big jump. Its earlier Muse Spark was a light, fast model that Wang himself called an "appetizer."
What this means for you
Muse Wonder shows where every big lab is heading. The selling points are not chat skills. They are memory, staying on task, and finishing multi-step work. That is the definition of agentic AI, and Meta now treats it as the main event.
It also shows how messy the choice has become. In one week, Meta alone shipped Muse Wonder, Muse Wonder Full, and Muse Spark Open, each with different access rules and prices. Picking the right model, plan, and API is now a job in itself.
You can skip that job. Our app, Agentic AI, gives you a team of ready-made expert agents on iPhone and iPad, like a Research Analyst, a Writing Coach, and a Code Helper. Give a task to one agent or let up to 4 work on it together. There is no account to create, and you get 3 free tasks to try it. Curious why teams work better? Read why two AI agents beat one.
Quick answers
When was Muse Wonder released? Meta released Muse Wonder and Muse Wonder Full on October 12, 2026.
Is Muse Wonder open source? No. Neither Muse Wonder nor Muse Wonder Full has downloadable weights. Meta’s open-weight model is Muse Spark Open.
What is the difference between Muse Wonder and Muse Wonder Full? Muse Wonder Full is the stronger model, but it only runs inside paid Muse Code plans. Muse Wonder is slightly weaker and is the one available through the Meta Model API.
How much does Muse Wonder cost? $4 per million input tokens and $20 per million output tokens on the Meta Model API. Muse Wonder Full comes with Muse Code plans at $50 or $200 a month.
Is Muse Wonder better than Claude Opus 5.5? Not on independent tests. Artificial Analysis scores Opus 5.5 at 58, Muse Wonder Full at 56, and Muse Wonder at 52. Meta says Wonder Full leads on some long coding benchmarks.
Muse Wonder vs Muse Spark: which is newer? Muse Wonder. Muse Spark came out in April 2026 as a light, fast model. Muse Wonder is Meta’s larger frontier model from October 2026.
Sources: Meta: Build with Muse Wonder · TechCrunch · VentureBeat · Futurum Group · Let’s Data Science
