🔍 Read the full analysis: Muse Wonder Full Brings Muse Wonder’s Charging Story To AI on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Meta has released Muse Wonder and Muse Wonder Full, two proprietary models with a one-million-token context window. Artificial Analysis’s October 12 evaluation gave Wonder Full a score of 56, while the API model scored 52; its estimated cost per benchmark task was $4.52, far above some similarly scoring rivals. Wonder Full is limited to Muse Code, and its per-task cost is not publicly available.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that received scores of 52 and 56, respectively, in an Artificial Analysis evaluation published October 12. The release gives Meta a model that scores near leading systems, but the stronger Wonder Full is available only through Muse Code, while the API-accessible Wonder had an estimated $4.52 cost per benchmark task in the evaluation.
Artificial Analysis said Muse Wonder Full scored 56 on its Intelligence Index v4.3, eight points above Muse Spark 1.3’s 48. The evaluation measured ten tasks, and the source report says Wonder Full improved on all ten and placed among the top three on six. Claude Opus 5.5 scored 58, while Claude Sonnet 5.5 also scored 56; those comparisons are reported scores, not evidence that the models perform identically in every use.
Muse Wonder is the version available through Meta’s API and inside Muse Code. Its listed price is $4 per million input tokens and $20 per million output tokens, with cached input listed at $0.20 per million tokens. Wonder Full is restricted to Muse Code users and has no public per-token price in the source material. Both models have a one-million-token context window, take text, images, audio and video as input, and return text.
Artificial Analysis estimated the cost of one Intelligence Index task at $4.52 for Muse Wonder. Its comparison put GPT-6.1 Sol at $0.72 for a score of 52, Gemini 4 Argon at $1.99 for 53, and Claude Sonnet 5.5 at $5.46 for 56. These are benchmark-task estimates, not a fixed charge for every real-world request; actual bills depend on usage and token counts.
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 Meet High API Costs
The release matters because it separates Meta’s strongest model from its purchasable one. Wonder Full’s score puts it near the top of the reported comparison, but organizations evaluating API providers cannot buy that model by the token based on the details provided. The public API option, Muse Wonder, scored four points lower and had a benchmark-task cost above several systems with similar or higher scores.
That difference affects decisions by developers and businesses weighing capability against operating cost. Artificial Analysis attributed much of Wonder’s estimated task bill to input: it said the model processed 7.69 million input tokens per task, with input accounting for $3.43, or 76%, of the $4.52 estimate. The report also gave Wonder Full an input volume of 6.44 million tokens per task, but no per-task price was available for that version.
The evaluation also records quality improvements that matter beyond a single overall score. The source report lists Muse Wonder Full at 57% on Terminal-Bench 4.0, up from 33% for Muse Spark 1.3, and says its hallucination rate on the cited measure fell from 33% to 23%. Those figures are encouraging performance signals, but the reported hallucination rate remains a reason for users to verify answers in workflows where errors carry costs.
As an affiliate, we earn on qualifying purchases.
How the Two Models Differ
The release follows Muse Spark 1.3 by six weeks, according to the source report. Artificial Analysis’s Intelligence Index comparison placed Spark 1.3 at 48, Muse Wonder at 52 and Wonder Full at 56. The source says Wonder Full improved on all ten evaluations, while Wonder’s listed results included 52% on Terminal-Bench, 68% on AutomationBench-AA and a 27% hallucination rate on the cited measure.
The models are not offered on equal terms. Wonder is available by API and in Meta’s coding product; Wonder Full is available only to Muse Code users, with no public token price described. Artificial Analysis also reported a score of 68 for the Muse Code plus Wonder Full configuration in its Coding Agents Index, matching its reported Claude Code plus Sonnet 5.5 result. That index assesses the model and its software harness together, so it does not establish how Wonder Full would perform in another coding environment.
Costs in the source are benchmark estimates, not universal prices for a task of a given type. The report’s comparison includes models with different token use and pricing, and the source material supplied here ends during its explanation of Muse Wonder’s input-cost breakdown. Readers should treat the supplied figures as the reported comparison, not as a complete audit of every model’s operating costs.
“The difference is the bill.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Pricing and Access Still Limited
Meta’s per-task price for Muse Wonder Full is not provided, and the source says it is available only to Muse Code users. It is not clear from the material whether Meta plans to offer it through an API, publish a token price, or expand access to other users.
The $4.52 figure for Muse Wonder comes from Artificial Analysis’s Intelligence Index task estimate. It should not be read as a guaranteed charge for an individual customer’s workload. The source material also cuts off partway through its detailed input-cost breakdown, leaving that explanation incomplete. The evaluation figures are attributed to Artificial Analysis, but the supplied material does not include its full methodology or confidence intervals for every listed comparison.
Finally, benchmark results do not settle how either model will perform on a particular company’s tasks. The reported hallucination rates and task scores describe specific evaluations; real-world accuracy, response times, availability and total cost for different workloads remain unreported here.
AI performance evaluation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Watch for Broader Model Access
The next practical milestone is whether Meta makes Muse Wonder Full available beyond Muse Code or discloses a price that lets customers compare its cost with API alternatives. No such expansion or pricing announcement appears in the supplied material. For now, API buyers can assess Muse Wonder using its stated token rates and the independent benchmark estimates, while Muse Code users can access the higher-scoring Full version under the reported restriction.
Users and procurement teams will also need to test the models against their own tasks. The Artificial Analysis scores offer a common point of comparison, but decisions about adoption depend on accuracy, token consumption and the cost of checking outputs in a specific workflow. Meta’s next disclosures on access, pricing and usage would clarify whether Wonder’s benchmark performance can translate into a competitive choice for developers.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Meta release?
Meta released Muse Wonder and Muse Wonder Full, proprietary models with a reported one-million-token context window. Both accept text, images, audio and video as input and return text.
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. The source report lists Muse Spark 1.3 at 48.
Can developers access Muse Wonder Full through an API?
The supplied report says Wonder Full is available only to Muse Code users. It does not provide a public API offering or per-token price for that model.
How much does Muse Wonder cost?
Meta’s listed rates in the report are $4 per million input tokens, $20 per million output tokens and $0.20 per million cached input tokens. Artificial Analysis estimated $4.52 to run one Intelligence Index task; that is a benchmark estimate, not a fixed price for every request.
What remains unknown?
The source does not say whether Meta will expand Wonder Full access or disclose a token price. It also does not establish how the benchmark results translate to individual customer workloads or what those workloads would cost.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
