🔍 Read the full analysis: Muse Wonder Full And Muse Wonder Bring New Energy To AI on ThorstenMeyerAI.com
Get hardware and tech essentials 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 an Intelligence Index score of 56, while the API model Muse Wonder scored 52. The evaluation estimates Muse Wonder costs $4.52 per benchmark task, compared with $0.72 for GPT-6.1 Sol at the same score; Wonder Full is available only through Muse Code, with no published per-token price in the source material.
Artificial Analysis reported that Muse Wonder Full scored 56, eight points above Meta’s previous model, Muse Spark 1.3, which scored 48. The evaluator said Wonder Full improved across all ten evaluations in the index and ranked in the top three on six. Among 27 model configurations included in the source’s comparison, Claude Opus 5.5 at its maximum setting scored higher, at 58, while Claude Sonnet 5.5 at its maximum setting also scored 56.
The gains included a rise on Terminal-Bench 4.0 from 33% for Muse Spark 1.3 to 57% for Wonder Full. Artificial Analysis also reported a GDPval-AA score of 1,857 Elo, placing Wonder Full between Opus 5.5 at 1,866 and Sonnet 5.5 at 1,838; the source says those results fall within the models’ confidence intervals. These are evaluator findings, not guarantees of performance in every user’s workflow.
Muse Wonder, the API model, scored 52 on the index. Artificial Analysis estimated $4.52 to run one index task with it, compared with $0.72 for GPT-6.1 Sol, also scored at 52. The source puts Gemini 4 Argon at 53 for $1.99 and Claude Sonnet 5.5 at 56 for $5.46 per task. These figures are benchmark-task estimates; they are not a universal measure of what every customer will pay.
Both models have a one-million-token context window and accept text, images, audio or speech, and video as input, returning text. Meta’s listed API prices for Wonder are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. The source reports that Wonder Full has no per-token price and is limited to Muse Code users.
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 gives Meta a stronger showing in independent benchmark comparisons, but the two models serve different audiences. Wonder Full’s higher score is confined to Muse Code, while developers seeking API access get the lower-scoring Wonder model at a listed price matching Claude Opus 5.5. The gap between the best-performing model and the model available by token makes access and pricing central to the announcement.
For buyers comparing providers, the reported $4.52 per-task estimate is a reason to examine actual usage patterns rather than scores alone. Artificial Analysis says Muse Wonder processes 7.69 million input tokens per index task, compared with 6.03 million for Muse Spark 1.3; the source attributes $3.43 of the estimated task cost to input. The model’s token use and the benchmark’s task design may not reflect a particular application, so customers would need to test their own workloads.
Quality also remains a practical consideration. Artificial Analysis reports a 23% hallucination rate for Wonder Full and 27% for Wonder, down from 33% for Spark 1.3, but higher than the 15% reported for Gemini 4 Argon among the compared models scoring at least 45 on the index. Benchmark improvements do not remove the need to verify outputs, particularly where errors could have financial or operational consequences.
As an affiliate, we earn on qualifying purchases.
From Spark 1.3 to Muse Wonder
The models arrived six weeks after Muse Spark 1.3, according to the source article. That short interval is relevant to the reported changes: Artificial Analysis recorded an eight-point Intelligence Index gain for Wonder Full and a 24-percentage-point improvement on Terminal-Bench 4.0. The evaluator describes Terminal-Bench as measuring whether models can complete work in a shell; the scores indicate a change in that test, not a general guarantee of agent performance.
The source also compares model-and-tool combinations in Artificial Analysis’s Coding Agents Index. Muse Code with Wonder Full scored 68, tying Claude Code with Sonnet 5.5 at 68. Claude Code with Opus 5.5 scored 66, while Muse Code with Wonder scored 63 and with Spark 1.3 scored 54. Because the index evaluates the model together with its harness, the result does not establish how Wonder Full would perform inside Claude Code.
Wonder’s published API rates are substantially higher than Spark 1.3’s listed rates in the source: $4 input and $20 output per million tokens, versus $1.25 and $4.25. The source says the new Wonder rates match Opus 5.5, while Sonnet 5.5, Gemini 4 Argon and GPT-6.1 Sol each list $2 per million input and $10 per million output tokens.
“The bill is in the input, not the output.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Pricing and Real-World Performance
The source material does not provide a per-token price for Muse Wonder Full, which is restricted to Muse Code, so its per-task cost cannot be compared on the same basis. It also does not include a response from Meta explaining the API pricing or the distinction between the models’ availability.
Artificial Analysis’s index and task-cost estimates are specific to its evaluation methods. The source does not establish how the models compare on every application, or whether the listed prices and token use will produce similar costs for customers’ own prompts. Confidence intervals also matter for some close benchmark comparisons, including the reported GDPval-AA results for Wonder Full and Anthropic’s models.
The reported hallucination rates are evaluator measurements, not a guarantee about error rates in other settings. It remains unclear from the source how the models perform across independent deployments, how often users will choose Wonder Full in Muse Code, and whether Meta will broaden access or publish additional pricing details.
As an affiliate, we earn on qualifying purchases.
API Use and Further Evaluations
Developers can access Muse Wonder through Meta’s API, while Muse Wonder Full is available to Muse Code users, according to the source. The next useful evidence will come from testing both models on real workloads, including the volume of input tokens required, the resulting API bills and the reliability of outputs under independent verification.
Readers and prospective users can also watch for further evaluation results or changes to availability and pricing. The source provides no timetable for additional releases, a wider rollout of Wonder Full, or publication of its per-token cost. Until more information appears, the benchmark scores and price estimates should be treated as a snapshot of the models and configurations assessed in October 2026.
As an affiliate, we earn on qualifying purchases.
Key Questions
What are Muse Wonder and Muse Wonder Full?
They are two proprietary Meta AI models released six weeks after Muse Spark 1.3, according to the source. Both accept text, images, audio or speech, and video as input, return text, and have a one-million-token context window.
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 reports that Wonder Full improved on all ten evaluations in the index.
Where can people use each model?
Muse Wonder is available through Meta’s API and inside Muse Code. Wonder Full is available only to Muse Code users, according to the 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 $4.52 to run one of its Intelligence Index tasks, but actual costs depend on usage and may differ by workload.
Does the benchmark establish which model is best for a particular task?
No. The reported scores reflect Artificial Analysis’s tests and configurations; they do not establish performance or cost for every real-world application. Users should evaluate relevant workloads and verify outputs, especially where errors could cause harm or financial loss.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
