🔍 Read the full analysis: Muse Wonder Full Joins The AI Frontier With A Muse Wonder Charge on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Artificial Analysis’s October 12 evaluation gave Meta’s Muse Wonder Full a score of 56 on its Intelligence Index, while API-accessible Muse Wonder scored 52. The results put Meta’s models near leading competitors, but the published task-cost estimates show Muse Wonder costing substantially more than some similarly scoring alternatives; Muse Wonder Full has no public per-token price.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that scored 52 and 56, respectively, in Artificial Analysis’s Intelligence Index v4.3 evaluation published October 12. The results place the Full version near leading models on the index, but the API-accessible Wonder model’s estimated cost per index task was $4.52, compared with $0.72 for a similarly scoring GPT-6.1 Sol configuration.
Artificial Analysis reported that Muse Wonder Full scored 56, eight points above Muse Spark 1.3’s 48, and improved on all ten evaluations included in the index. Among 27 configurations compared in the source report, Claude Opus 5.5 at its maximum setting scored higher, with 58; Claude Sonnet 5.5 at its maximum setting also scored 56. Muse Wonder scored 52, matching GPT-6.1 Sol at its maximum setting.
The two models share a one-million-token context window and accept text, images, audio and video as input, returning text. Meta makes Muse Wonder available through its API and within Muse Code, while Muse Wonder Full is limited to Muse Code. Muse Wonder’s listed rates are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. The source report says Muse Wonder Full has no per-token price, leaving its cost per task unavailable for comparison.
Performance varied by task. Muse Wonder Full scored 57% on Terminal-Bench 4.0, compared with 33% for Muse Spark 1.3, and ranked among the top three on six of the ten index evaluations. Its score on GDPval-AA, a measure of economically useful knowledge work, was 1,857 Elo, near Claude Opus 5.5’s 1,866 and Claude Sonnet 5.5’s 1,838. The report notes those results fall within the models’ confidence intervals. These are benchmark findings, not evidence by themselves of performance in every real-world deployment.
Infographic 01 · Where Meta’s new pair lands
Muse Wonder Full ties for second. Muse Wonder ties GPT-6.1 Sol.
Artificial Analysis scores 27 frontier and open-weight configurations on ten evaluations. Meta’s two new models land near the top of the list. Six weeks earlier, Muse Spark 1.3 scored 48.
Intelligence Index score, 27 configurations
Higher is better. Reasoning effort in brackets, as Artificial Analysis tested it.
Availability: Muse Wonder runs on Meta’s API and in Muse Code. Muse Wonder Full is only available to Muse Code users and has no per-token price.
Source: Artificial Analysis, Muse Wonder analysis (12 Oct 2026) and model leaderboard, read 13 Oct 2026. Intelligence Index v4.3, ten evaluations.
Powered by Thorsten Meyer AI
Strong Scores, Uneven Value
The release gives businesses and developers a new option from Meta with benchmark results close to leading proprietary systems. In Artificial Analysis’s Coding Agents Index, Muse Code with Muse Wonder Full scored 68, tying the listed Claude Code and Claude Sonnet 5.5 configuration. That index evaluates the model together with its software harness, so the result applies to that combined setup; it does not establish how Wonder Full would perform inside Claude Code.
For buyers, however, the performance score is only part of the decision. Artificial Analysis estimated that one Intelligence Index task cost $4.52 with Muse Wonder, against $0.72 for GPT-6.1 Sol, which also scored 52, and $1.99 for Gemini 4 Argon, which scored 53. These are task-cost estimates from the evaluation, not a universal bill for every workload. Still, they point to a practical issue: model performance must be weighed against token use and pricing, especially for repeated or high-volume jobs.
The report also gives reason for care in workflows where errors carry costs. Muse Wonder Full had a reported hallucination rate of 23%, down from 33% for Spark 1.3, while Muse Wonder was at 27%. Those figures are specific to the source’s evaluation method. They do not mean the models will produce errors at those exact rates in every use case, but they support retaining review and verification for consequential outputs.
As an affiliate, we earn on qualifying purchases.
From Spark to Wonder
The release follows Muse Spark 1.3 by about six weeks, according to the source report. In that interval, Meta’s new Full model rose from Spark’s 48 to 56 on the Artificial Analysis Intelligence Index, while the API model reached 52. The source attributes the scores and comparative results to Artificial Analysis’s independent evaluation published October 12, 2026.
Meta has separated access to the two models: Muse Wonder can be called through the API and used in Muse Code, while Wonder Full is available only in Muse Code. That means developers can compare the API model’s listed rates with alternatives, but the more highly rated version does not have a public per-token price in the supplied material. The release therefore offers different trade-offs in capability, availability and cost transparency.
Token consumption is relevant to the price comparison. Artificial Analysis’s figures show Muse Wonder using about 7.69 million input tokens per index task, versus 6.03 million for Spark 1.3. The report says Wonder’s input use was higher than that of several compared models, including Argon, Sol and Astra. It estimates input tokens account for $3.43, or 76%, of Wonder’s $4.52 task cost. These results describe the benchmark workload and should not be treated as a forecast for every customer’s usage.
“The difference is the bill.”
— Thorsten Meyer, ThorstenMeyerAI.com
AI model performance benchmarking tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Full Model Cost Remains Unknown
The source material does not provide a per-token price for Muse Wonder Full, which is available only through Muse Code. Its cost per benchmark task, and how that cost compares with other coding-agent configurations, therefore cannot be calculated from the information supplied. Meta’s pricing for the Full model and any usage limits or eligibility conditions are also not specified there.
It is also unclear how closely the benchmark task costs will match customer bills. The estimates reflect Artificial Analysis’s evaluation workload and token counts; actual costs can vary with prompts, outputs, caching and how a model is used. The report does not provide enough information to infer the price of other workloads from the $4.52 figure. Benchmark scores likewise do not establish performance across every industry or production setting.
The reported hallucination rates and task scores depend on evaluation design and comparison conditions. The source gives no further details here about confidence intervals for every individual result, or about independent tests beyond those cited. Buyers will need to judge the models against their own tasks and reliability requirements.
cost-effective AI text generation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Pricing and Real-World Use
The next useful comparison will be between published pricing for Muse Wonder Full and its measured performance on customer workloads. Meta’s access and pricing details could clarify whether the stronger model is practical beyond Muse Code and how its total cost compares with alternatives. No further release date or pricing announcement is included in the source material.
For now, developers considering Muse Wonder can compare its listed API rates with the cost estimates in the Artificial Analysis evaluation, while testing token use and output quality on representative tasks. Organizations using either model for consequential work should keep human checks or other verification suited to the risk, since the reported hallucination rates indicate that confident errors remain possible.
AI input-output token management tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What are Muse Wonder and Muse Wonder Full?
They are Meta’s two newly released proprietary AI models. Both accept text, images, audio and video as input, return text, and have a one-million-token context window, according to the source report.
How did the models score?
Artificial Analysis’s Intelligence Index v4.3 gave Muse Wonder Full a score of 56 and Muse Wonder a score of 52. Muse Spark 1.3 scored 48 in the comparison cited by the report.
Where can people use each model?
Muse Wonder is available through Meta’s API and Muse Code. Muse Wonder Full is available only in Muse Code, according to the supplied source material.
How much does Muse Wonder cost?
Meta’s listed rates are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimated a $4.52 cost for one Intelligence Index task; that is an evaluation-specific estimate, not a fixed price for every task.
Is Muse Wonder Full more expensive than Muse Wonder?
The supplied material does not give a per-token price for Muse Wonder Full. Because it is limited to Muse Code, its cost per task cannot be compared from the information provided.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
