🔍 Read the full analysis: Muse Wonder Full Joins AI’s Frontier; Muse Wonder Charges Along 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-available Muse Wonder scored 52. The evaluation also estimated a $4.52 cost per Intelligence Index task for Muse Wonder, compared with $0.72 for GPT-6.1 Sol at the same score; Meta has not published a per-token price for Wonder Full.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that differ in performance and access, according to a new evaluation by Artificial Analysis. Published October 12, the evaluation scores Wonder Full at 56 on the Intelligence Index and the API model Wonder at 52; it estimates that one benchmark task costs $4.52 on Wonder, versus $0.72 on GPT-6.1 Sol, which received the same score.
Artificial Analysis says Muse Wonder Full improved on all ten evaluations in its Intelligence Index v4.3 compared with Muse Spark 1.3, rising from 48 to 56. Among 27 configurations included in the source comparison, only Claude Opus 5.5 in its max configuration scored higher, at 58; Claude Sonnet 5.5 in max also scored 56. These are benchmark results, not a guarantee of performance on every task.
The two models are reported to accept text, images, audio and video and to have a one-million-token context window. Wonder is available through Meta’s API and in Muse Code, Meta’s coding agent. The source gives its list price as $4 per million input tokens and $20 per million output tokens, with cached input priced at $0.20 per million tokens. Wonder Full is available only through Muse Code and has no published per-token price in the source, so its API-style cost cannot be compared directly.
Artificial Analysis reports that Wonder Full scored 57% on Terminal-Bench 4.0, up from 33% for Spark 1.3, and placed among the top three on six of the ten index evaluations. Wonder scored 52% on Terminal-Bench and 68% on AutomationBench-AA, compared with 58% for Spark 1.3. The evaluation also reports hallucination rates of 23% for Wonder Full and 27% for Wonder, down from 33% for Spark 1.3.
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
Benchmark Gains Meet API Costs
The release presents two different questions for users: how capable the models are and what it costs to use the model available through the API. Artificial Analysis’s results place Wonder Full near the top of its measured group, while the API model’s score of 52 matches GPT-6.1 Sol in the source comparison. Its estimated task cost, however, is substantially higher.
For the Intelligence Index task estimate, Artificial Analysis lists $4.52 for Muse Wonder, $0.72 for GPT-6.1 Sol at a score of 52, and $1.99 for Gemini 4 Argon at a score of 53. The comparison is specific to the evaluation’s task and pricing assumptions; real-world costs can vary with prompts, usage patterns and billing terms. Still, the gap may matter to businesses deciding whether to test or deploy the API model at scale.
The same evaluation attributes much of Wonder’s estimated bill to input: it says the model consumes 7.69 million input tokens per task, accounting for $3.43, or 76% of the estimated $4.52. That is more input than the listed figures for several competitors. Token use and benchmark results can help frame a cost assessment, but neither alone establishes which model is the best fit for a particular workflow.
As an affiliate, we earn on qualifying purchases.
Two Models, Different Access
The models arrive six weeks after Muse Spark 1.3. In the source comparison, Spark scored 48 on the Intelligence Index, while Wonder Full scored 56 and Wonder scored 52. The source also reports that Wonder’s list prices are higher than Spark’s: Spark was listed at $1.25 per million input tokens and $4.25 per million output tokens, compared with Wonder’s $4 and $20.
Artificial Analysis also evaluated model-and-tool combinations in its Coding Agents Index. It gave Muse Code with Wonder Full a score of 68, matching Claude Code with Sonnet 5.5 in the cited results. Muse Code with Wonder scored 63, while the Spark 1.3 configuration scored 54. Because this index measures a model together with its coding harness, those scores describe the combined systems, not the models operating independently or in another tool.
The source’s benchmark findings are attributed to Artificial Analysis, which published the evaluation on October 12. They should be read as results under that evaluator’s methods and selected configurations. The article does not provide enough information to independently verify the testing setup, confidence intervals for every comparison, or the commercial terms users may receive outside listed prices.
multimodal AI models for text image audio video
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Pricing and Performance Caveats
Muse Wonder Full’s per-token or per-task cost is not disclosed in the source, limiting direct cost comparisons with API models. It is also unclear whether Meta plans to make Wonder Full available through its API or publish additional pricing details.
The reported task costs are estimates based on Artificial Analysis’s benchmark, token use and listed prices. The source does not establish that the same cost would apply to a typical customer’s workload. Nor do benchmark scores settle how the models compare across different prompts, software environments or tasks. The source reports confidence-interval context for some GDPval-AA results but does not give comparable uncertainty details for every score.
The reported decline in hallucination rates is progress within the cited evaluation, not proof that either model will avoid confident errors in practical use. For decisions where incorrect output carries a cost, the evaluation alone does not remove the need for appropriate verification.
As an affiliate, we earn on qualifying purchases.
Watch for Access and Pricing
The next developments to watch are whether Meta makes Wonder Full available beyond Muse Code, whether it publishes a clear pricing schedule for that model, and whether further independent evaluations reproduce the reported results. Meta’s API availability and current list prices for Wonder are described in the source, but no additional rollout timetable is provided.
Users comparing the models can assess benchmark results alongside their own task requirements, token consumption and billing terms. For organizations considering deployment, the key unresolved practical question is whether Wonder’s cost and performance on their workloads justify its use compared with lower-priced alternatives. The available evaluation offers one comparison point, not a forecast of future performance or spending.
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. The source says both accept text, images, audio and video, and have a one-million-token context window. Wonder is available through Meta’s API and Muse Code; Wonder Full is available only through Muse Code.
How did the models score?
Artificial Analysis scored Muse Wonder Full at 56 and Muse Wonder at 52 on its Intelligence Index v4.3. The source reports a score of 48 for Muse Spark 1.3 in the same comparison.
How much does Muse Wonder cost?
The source lists Wonder at $4 per million input tokens, $20 per million output tokens, and $0.20 per million cached input tokens. Artificial Analysis estimated $4.52 for one of its Intelligence Index tasks; that estimate is not necessarily the cost of a typical user workload.
Is Muse Wonder Full available through the API?
The source says Wonder Full is available only to Muse Code users. It provides no per-token price or confirmed timetable for wider availability.
Do the benchmark results prove the models are reliable?
No. The results describe performance in Artificial Analysis’s evaluations, not every real-world task. The source reports hallucination rates of 23% for Wonder Full and 27% for Wonder, so users should account for the possibility of incorrect output.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
