🔍 Read the full analysis: Muse Wonder Full Joins AI’s Frontier As Muse Wonder Gets A Charge on ThorstenMeyerAI.com
Get the little things that make your day 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
Meta has released Muse Wonder and Muse Wonder Full, two proprietary models that Artificial Analysis evaluated on October 12, 2026. Wonder Full scored 56 on the Intelligence Index, while the API-available Wonder scored 52 and cost $4.52 per benchmark task—far more than some similarly scoring rivals. Wonder Full is available only through Muse Code, and its per-task cost is not disclosed.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary models that earned strong results in an October 12 Artificial Analysis evaluation but differ in access and price. Wonder Full scored 56 on the Intelligence Index and is available only through Meta’s Muse Code; Wonder scored 52 and is available through Meta’s API, where Artificial Analysis put the cost of one index task at $4.52.
Artificial Analysis rated Muse Wonder Full eight points above Muse Spark 1.3, which scored 48, and reported improvement on all ten evaluations in the index. Among 27 configurations compared in the source report, only Claude Opus 5.5 in its max configuration scored higher, at 58; Claude Sonnet 5.5 max tied Wonder Full at 56. These are benchmark results, not a guarantee of performance in every user’s tasks.
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 via its API and within Muse Code, its coding agent. Its listed prices are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Wonder Full is limited to Muse Code and has no per-token price, leaving its cost outside the API comparison.
Artificial Analysis also reported a sharp rise in Wonder Full’s Terminal-Bench 4.0 result, from 33% for Spark 1.3 to 57%. The benchmark tests a model’s ability to complete tasks in a shell environment. Wonder Full placed in the top three on six of the ten evaluations, according to the source report. In the Coding Agents Index, Muse Code paired with Wonder Full scored 68, matching Claude Code with Sonnet 5.5; the index measures the model and its software harness together.
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 Access and Cost
The release puts Meta in a stronger position in independent comparisons of advanced models and coding agents. Wonder Full’s 56 score and its coding-agent result show that Meta’s own model-and-tool combination can match a leading configuration in this particular index. That gives developers using Muse Code a new option, though it does not establish that Wonder Full would perform the same way in another harness or on every coding task.
The clearest decision point for customers is the gap between benchmark performance and API economics. Artificial Analysis estimated $4.52 to run one Intelligence Index task with Muse Wonder. It estimated $0.72 for GPT-6.1 Sol, which also scored 52; Gemini 4 Argon scored 53 at $1.99. These figures describe the benchmark’s task costs, not a universal charge for every real-world request, but they make the API model’s relative expense relevant to organizations weighing repeated use.
Access also limits what customers can compare directly. The stronger-scoring Wonder Full is confined to Muse Code, and its per-token or per-task cost is not available in the supplied material. The API model is easier to integrate and price, but its score is lower than Full’s and its benchmark task cost is higher than several named alternatives. Customers therefore cannot infer that Meta’s highest benchmark result is also its most accessible or cost-effective offering.
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. The prior model scored 48 on the Intelligence Index. Muse Wonder’s 52 and Wonder Full’s 56 mark separate steps in Meta’s latest lineup, but the models do not share the same distribution: one is an API product and the other is available only inside Muse Code.
The source compares Muse Wonder’s $4.52 benchmark task cost with $1.60 for Spark 1.3. It attributes much of Wonder’s bill to input processing: the model used 7.69 million input tokens per task, compared with 6.03 million for Spark 1.3. Artificial Analysis put input at $3.43, or 76% of Wonder’s estimated task cost. Wonder generated about 54,000 output tokens per task, fewer than some tested models, so the report identifies input-token use as the larger contributor in this comparison.
Scores and prices come from the source’s account of Artificial Analysis testing; they should be read with their stated configurations and benchmark method in mind. The Coding Agents Index combines the model with its harness, while the Intelligence Index compares model configurations. Neither result alone settles which system will be best for a particular workflow.
“The stronger model is the one you can’t buy by the token, and the model you can buy competes one tier lower.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Costs and Real-World Performance
The supplied material does not give a per-token price or benchmark task cost for Muse Wonder Full, so its economics cannot be compared directly with the API models. It also does not specify when, or whether, Meta will make Full available outside Muse Code. The figures for Muse Wonder describe one Intelligence Index task; the report does not establish that the same cost will apply to customers’ workloads.
Benchmark scores do not resolve how the models will perform across different production tasks, prompting approaches or software environments. The source reports hallucination rates of 23% for Wonder Full and 27% for Wonder, compared with 33% for Spark 1.3, but does not provide enough detail here to treat those rates as universal error probabilities. The results suggest improvement on the named measure, not that users can dispense with verification.
As an affiliate, we earn on qualifying purchases.
Pricing and Broader Availability
The immediate next step for buyers is to assess Muse Wonder through its stated API pricing and test it against their own workloads, paying attention to input-token consumption as well as output. Organizations evaluating Muse Code can compare Wonder Full within that environment, but the supplied report does not state a separate price for using the model there.
Further clarity would come from Meta publishing Wonder Full’s costs, expanding its availability, or providing more information about how the models perform in customer deployments. Until then, the published evidence is limited to the named benchmark results and access routes: a high-scoring, Muse Code-only Full model and an API model with a lower score and comparatively high measured task cost.
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 one-million-token context window. They accept text, images, audio and video as input and return text.
How did Muse Wonder Full score?
Artificial Analysis gave Muse Wonder Full a score of 56 on its Intelligence Index. In the configurations cited by the source report, Claude Opus 5.5 max scored 58 and Claude Sonnet 5.5 max also scored 56.
Can customers use Muse Wonder Full through Meta’s API?
The supplied report says Wonder Full is available only to Muse Code users. Muse Wonder, the separately named model, is available through Meta’s API and inside Muse Code.
How much does Muse Wonder cost?
Meta’s listed API prices are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimated $4.52 for one Intelligence Index task; that benchmark estimate is not a fixed price for every user request.
What remains unknown about the release?
The supplied material gives no per-token price or task-cost estimate for Muse Wonder Full and does not say whether it will become available outside Muse Code. Real-world costs and performance will also depend on customers’ workloads and usage.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
