Meta’s Muse Wonder Matches Claude Fable for Less

Set Trending Topics as a preferred source on Google.

More brains, less make-believe, a bigger bill: The independent benchmarkers at Artificial Analysis have just taken a detailed look at Meta’s new A.I. models, Muse Wonder and Muse Wonder Full. Their verdict: a “material jump” over previous Meta models. In the Intelligence Index, Muse Wonder Full even overtakes Anthropic’s Claude Fable 5.1, and the standard Muse Wonder trails it by just one point. Yet Muse Wonder costs about 40 percent less than Fable per test task.

Muse Wonder Full scores 56 points in the Intelligence Index, and the standard Muse Wonder scores 52. Its predecessor, Muse Spark 1.3, scored 48, even though only six weeks separate the releases. Trending Topics has already reported on how Meta overtook GPT-6 and Gemini 4 with the new models.

The Intelligence Index Top 10

This is what the top of the Artificial Analysis ranking currently looks like (Intelligence Index points, each at the highest reasoning setting):

1. Claude Opus 5.5 (Anthropic): 57.6
2. Claude Sonnet 5.5 (Anthropic): 56.0
3. Muse Wonder Full (Meta): 55.8
4. Claude Fable 5.1 (Anthropic): 53.4
5. GPT-6 Astra (OpenAI): 52.7
6. Gemini 4 Argon (Google): 52.6
7. Muse Wonder (Meta): 52.4
8. GPT-6.1 Sol (OpenAI): 51.8
9. Muse Spark 1.3 (Meta): 48.1
10. Grok 4.7 (xAI): 46.4

The Price Ranking

On price, the picture partly flips. Artificial Analysis calculates what a model costs on average per Intelligence Index task. That depends not only on the price per token but also on how many tokens a model uses for a task. The Intelligence Index top 10, from most to least expensive (list prices per million input and output tokens in parentheses):

1. Claude Fable 5.1: $7.63, about €6.80 ($10 / $50)
2. Claude Opus 5.5: $5.98, about €5.30 ($4 / $20)
3. Claude Sonnet 5.5: $5.46, about €4.90 ($2 / $10)
4. Muse Wonder: $4.52, about €4 ($4 / $20)
5. Grok 4.7: $3.74, about €3.30 ($2 / $6)
6. GPT-6 Astra: $3.26, about €2.90 ($10 / $50)
7. Gemini 4 Argon: $1.99, about €1.80 ($2 / $10)
8. Muse Spark 1.3: $1.60, about €1.40 ($1.25 / $4.25)
9. GPT-6.1 Sol: $0.72, about €0.60 ($2 / $10)

Muse Wonder Full is missing from this ranking because Meta does not charge for it per token.

The Strengths

• A.I. agents: The biggest gains come in tasks that A.I. agents complete on their own. On Terminal-Bench 4.0, Muse Wonder Full jumps from 33 to 57 percent, putting it level with Google’s Gemini 4 Argon.
• Knowledge work and science: Muse Wonder Full places in the top three on six of the index’s ten tests. On CritPt (scientific reasoning) and GDP.pdf (knowledge work with documents), it even ranks first.
• Coding: In the Coding Agents Index, Muse Wonder Full scores 68 points in Meta’s own coding environment, Muse Code. That matches Claude Code with Claude Sonnet 5.5; Claude Opus 5.5 scores 66 there.
• Few hallucinations: Muse Wonder Full makes up facts far less often than its predecessor, with its hallucination rate falling from 33 to 23 percent. Among the top models, only Gemini 4 Argon does better, at 15 percent.
• Leaner output: Both models use about 10 percent fewer output tokens per task than Muse Spark 1.3, and also fewer than Gemini 4 Argon or Claude Fable 5.1.

The models also offer a context window of one million tokens and can process text, images, video and audio. They respond only in text, however.

The Weaknesses

• Pricier than its predecessor: Meta has raised prices sharply. Instead of $1.25 and $4.25, a million input and output tokens for Muse Wonder now cost $4 and $20 (about €3.60 and €17.80). That makes Muse Wonder 2.8 times as expensive per test task as Muse Spark 1.3. It is cheaper than the Claude models, but GPT-6.1 Sol and Gemini 4 Argon deliver comparable performance for far less.
• A big appetite for input: Muse Wonder processes 7.69 million input tokens per task, two to seven times as many as the models from Google and OpenAI. About three-quarters of its cost comes from that input.
• Gaps in knowledge: On the AA-Omniscience knowledge test, Muse Wonder Full scores only 30 points and the standard version 25. Claude Opus 5.5 scores 46, GPT-6 Astra 43 and Gemini 4 Argon 42.
• Limited access: The stronger Muse Wonder Full is currently available only to Muse Code users. Only the standard version is offered through Meta’s application programming interface (API).

Narrow Margins at the Top

The lead is not quite as clear as the rankings suggest. The leads on CritPt and GDP.pdf are “narrow,” Artificial Analysis writes, with the nearest competitors at 31 to 32 percent and Muse Wonder Full at 33 percent. On GDPval-AA, a test in which A.I. models complete real office tasks, Muse Wonder Full’s result falls within the statistical margin of error of Claude Opus 5.5 and Claude Sonnet 5.5.