🔍 Read the full analysis: Muse Wonder Full Brings A New AI Frontier Story, And Muse Wonder Charges on ThorstenMeyerAI.com
TL;DR
Meta has released Muse Wonder and Muse Wonder Full, two proprietary models with a one-million-token context window. Artificial Analysis scored Wonder Full at 56 and the API-available Wonder at 52, while estimating a $4.52 cost per Intelligence Index task for Wonder—far above some models with similar scores. Wonder Full is available only through Muse Code, and its per-task cost is not public.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that scored 52 and 56, respectively, on the Artificial Analysis Intelligence Index v4.3. The stronger model is available only through Meta’s Muse Code coding agent; the API-accessible Wonder costs an estimated $4.52 per benchmark task, according to Artificial Analysis, making its price a central point of comparison with similarly scoring rivals.
Artificial Analysis published its independent evaluation on October 12, 2026, six weeks after Meta’s Muse Spark 1.3 release. It scored Wonder Full at 56, up from Spark 1.3’s 48, and reported improvement across all ten evaluations in the index. Among the 27 configurations cited in the source report, only Claude Opus 5.5 at its maximum setting scored higher, with 58; Claude Sonnet 5.5 at maximum also scored 56. These are benchmark results, not a measure of performance in every real-world use.
The two models share a one-million-token context window and accept text, images, audio or speech, and video as input, returning text. Meta offers Muse Wonder through its API and Muse Code. Its listed prices are $4 per million input tokens, $0.20 per million cached input tokens, and $20 per million output tokens. Wonder Full has no public per-token price and is available only to Muse Code users, according to the source material.
On the same index, Muse Wonder scored 52, matching GPT-6.1 Sol’s cited score. Artificial Analysis estimated that one index task cost $4.52 with Wonder, compared with $0.72 for Sol, $1.99 for Gemini 4 Argon, and $3.26 for GPT-6 Astra. Those figures are estimates tied to the benchmark’s task and usage, not a universal charge for every user request. The source report says Wonder used 7.69 million input tokens per task on average, accounting for most of the estimated bill.
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, Higher API Costs
The release puts Meta into closer competition with leading models on independent benchmarks, while raising a practical question for developers: what performance does the API price buy? Wonder Full’s score of 56 places it near the cited Claude models on the Intelligence Index, and the Muse Code plus Wonder Full configuration scored 68 on Artificial Analysis’ Coding Agents Index, tying its cited Claude Code plus Sonnet 5.5 configuration.
That coding-agent result applies to the model-and-harness combination, not to Wonder Full in every coding environment. It also does not establish that the model would score the same when run through a different tool setup. For organizations comparing services, Wonder’s estimated task cost is a distinct consideration from its benchmark score. The reported $4.52 is much higher than Sol’s $0.72 at the same index score, though actual expenses will depend on workload, token use, and pricing conditions.
Reliability remains relevant alongside capability and cost. Artificial Analysis put Wonder’s hallucination rate at 27% and Wonder Full’s at 23%, compared with 33% for Muse Spark 1.3 in the report. The improvement is a benchmark finding, not a guarantee that a model will avoid errors in deployment. The source also reports stronger results from some other models on its knowledge and hallucination measure, so users handling consequential work still have reason to verify outputs.
As an affiliate, we earn on qualifying purchases.
From Muse Spark to Wonder
Muse Wonder and Wonder Full follow Muse Spark 1.3 by six weeks. In the Artificial Analysis evaluation cited by the source, Spark 1.3 scored 48 on the Intelligence Index. Wonder Full reached 56, while the API model Wonder scored 52. The source report also lists Wonder Full’s Terminal-Bench result at 57%, compared with 33% for Spark 1.3, and Wonder’s at 52%.
Artificial Analysis’ Intelligence Index combines results across ten evaluations, so an overall score can conceal differences between tasks. The source says Wonder Full ranked in the top three on six of those evaluations. On GDPval-AA, a measure of economically useful knowledge work, it scored 1,857 Elo; the report gives Claude Opus 5.5 as 1,866 and Sonnet 5.5 as 1,838, while noting the results fall within the models’ confidence intervals. Such close scores should not be read as a definitive ranking.
The separate Coding Agents Index measures models together with their agent software. The source gives Muse Code plus Wonder Full a score of 68, the same as Claude Code plus Sonnet 5.5; Claude Code plus Opus 5.5 scored 66. The figures provide a comparison of those specific configurations, rather than a direct test of the models alone.
„The difference is the bill.“
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Public Pricing and Real-World Results
Meta’s public per-token pricing for Muse Wonder Full is not provided in the source material because the model is available only through Muse Code. Its cost per benchmark task, and how that compares with Wonder’s API cost, therefore cannot be calculated from the reported information. The source does not include a statement from Meta explaining the pricing or access decision.
The $4.52 estimate for Wonder reflects Artificial Analysis’ benchmark task and observed token consumption. It is not a fixed price for every prompt, and the source does not establish that customers will use the same number of tokens across different applications. Benchmark scores also do not settle how either model will perform across individual businesses, coding environments, or safety requirements. The article cites evaluation results, but provides no details here about sample sizes or all uncertainty ranges beyond the stated GDPval-AA comparison.
As an affiliate, we earn on qualifying purchases.
What Developers Can Compare
Developers can assess Muse Wonder through Meta’s API or test Wonder Full within Muse Code, subject to each product’s availability and terms. Practical comparisons should track task quality, token consumption, latency, and total cost on representative workloads rather than assume that a benchmark score predicts a production bill. For high-stakes use, the reported hallucination figures also support retaining human review or other checks.
The next information that would clarify the launch is Meta’s explanation of Wonder’s pricing and any public cost details or broader access for Wonder Full. Until that is available, the benchmark results provide an independent performance snapshot, while the cost of using the stronger, Muse Code-only model remains unknown.
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. Both accept text, images, audio or speech, 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 scored Muse Wonder Full at 56 and Muse Wonder at 52 on its Intelligence Index v4.3. The source says Wonder Full improved on all ten index evaluations; scores do not guarantee the same results on every task.
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 to run one of its Intelligence Index tasks with Wonder; that task-based estimate is not a fixed charge for every request.
Can anyone access Muse Wonder Full through the API?
Not according to the source material. It says Wonder Full is available only to Muse Code users and has no public per-token price. The report does not specify whether Meta plans wider access.
Does the benchmark show Wonder is more expensive than every rival?
No. The cited comparison covers particular models and estimated costs for one Intelligence Index task. It reports that Wonder’s $4.52 estimate exceeds the estimates for GPT-6.1 Sol, Gemini 4 Argon, and GPT-6 Astra, but costs vary with token use and workload.
Source: ThorstenMeyerAI.com
