Meta's Muse Wonder Full ranks second on Artificial Analysis index behind Claude Opus 5.5
In brief
- Meta released Muse Wonder and Muse Wonder Full on October 12, 2026.
- Muse Wonder Full scored 56 on the Artificial Analysis Intelligence Index, second behind Claude Opus 5.5.
- Standard Muse Wonder scored 52, versus 48 for Meta's earlier Muse Spark 1.3.
- Muse Wonder pricing: $4 per million input tokens and $20 per million output tokens.
- Muse Wonder Full, the 56-point model, is available only to subscribers.
Where the scores land
On the Artificial Analysis index, the standard Muse Wonder scored 52 and Muse Wonder Full reached 56 at max effort, according to Crypto Briefing. Meta's previous model, Muse Spark 1.3, scored 48 on the same index. That's an 8-point jump for the Full version.
Muse Wonder Full also placed in the top three on six of the ten evaluations that make up the index. It did best on agentic tasks and coding benchmarks.
It still isn't first.
Claude Opus 5.5 holds the top spot on the aggregate Artificial Analysis score. The post announcing the results tells a different story: it said Muse Wonder beats Claude Opus 5.5 on most benchmarks. According to the same announcement, Muse Wonder also beats OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1 on coding, legal and automation tests. These are Meta's own comparisons. The aggregate ranking doesn't back the "most benchmarks" claim.
Pricing, access and the subscriber split
Meta's announcement, as reported by Crypto Briefing, said Muse Wonder is available through the Meta Model API and Muse Code. Muse Wonder Full is restricted to subscribers. Meta set the price of Muse Wonder at $4 per million input tokens and $20 per million output tokens.
The split matters to developers. The model that scored 56 is the subscriber-only one, so developers calling the API at those per-token prices get the model that scored 52.
Meta also released the models with two features, Active Memory and Active Focus. The company said both are meant to make the models more dependable during agentic tasks (the area where the Full model scored best).
Meta's pitch against OpenAI and Anthropic
Meta is positioning the Muse Wonder family as a direct challenge to the current AI leaders, particularly OpenAI and Anthropic. In its own analysis, Crypto Briefing said the release shows how quickly Meta has sped up its AI development since a 2025 overhaul of that work.
For now, Meta's best index score is in second place, and it's only available to subscribers.
Frequently asked questions
Is Meta's Muse Wonder Full the top-ranked model on the Artificial Analysis index?
No. Muse Wonder Full scored 56 on the Artificial Analysis Intelligence Index v4.3 at maximum reasoning effort, which made it the second-highest-rated model. Anthropic's Claude Opus 5.5 still holds the top spot on the aggregate score.
How much does Muse Wonder cost, and who can access Muse Wonder Full?
Meta priced Muse Wonder at $4 per million input tokens and $20 per million output tokens. It's available through the Meta Model API and Muse Code. Muse Wonder Full, the version that scored 56, is restricted to subscribers.
