🔍 Read the full analysis: How Muse Wonder Full Enters The AI Frontier While Muse Wonder Charges 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 1-million-token context window. Artificial Analysis scored the Muse Code-only Full version at 56 and the API model at 52; its evaluation estimates Muse Wonder costs $4.52 per Intelligence Index task, compared with $0.72 for GPT-6.1 Sol at the same score.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that Artificial Analysis evaluated on October 12, 2026. The evaluator gave Muse Wonder Full an Intelligence Index score of 56, while the API-available Muse Wonder scored 52; its estimated cost per benchmark task was $4.52, compared with $0.72 for GPT-6.1 Sol at the same score.
Artificial Analysis says Muse Wonder Full improved on all 10 evaluations in its Intelligence Index v4.3, placing among the top three on six. Its score of 56 is eight points above Muse Spark 1.3’s 48. Among 27 configurations compared in the source report, only Claude Opus 5.5 at 58 scored higher; Claude Sonnet 5.5 also scored 56. These are benchmark results, not a guarantee of equivalent performance on every task.
The strongest model has narrower access: Muse Wonder Full is available only inside Muse Code, Meta’s coding agent, and has no published per-token price. Muse Wonder is available through Meta’s API and in Muse Code. Both models have a 1-million-token context window, accept text, images, audio and video, and return text, according to the source report.
Muse Wonder’s listed API rates are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. The report says those rates match Claude Opus 5.5’s list price. Artificial Analysis estimates a Muse Wonder Intelligence Index task at $4.52, versus $0.72 for GPT-6.1 Sol, $1.99 for Gemini 4 Argon and $5.46 for Claude Sonnet 5.5. The costs are estimates tied to the evaluator’s benchmark workload, not fixed charges for all user tasks.
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
A Stronger Model, With Limited Access
The release puts Meta closer to leading models on several reported tests, but the two versions create a practical divide: the higher-scoring model is confined to Muse Code, while the version customers can call through the API scores lower and is priced at the same list rates as Opus 5.5. The source report does not provide a per-task price for Muse Wonder Full, so customers cannot directly compare its benchmark cost with the API model.
For buyers choosing an API, the cost comparison is material. Artificial Analysis reports that Muse Wonder and GPT-6.1 Sol both score 52 on its index, while its modeled task cost for Muse Wonder is about 6.3 times Sol’s. That does not establish which model is cheaper for a particular business workflow: token use and task requirements vary. It does give teams a reason to test their own workloads before treating benchmark scores or list prices as a purchasing decision.
The report also flags reliability as a consideration. Muse Wonder Full’s reported hallucination rate is 23%, down from 33% for Muse Spark 1.3, while Muse Wonder’s is 27%. The evaluator says Gemini 4 Argon’s rate is 15% among models scoring at least 45 on the index. These measurements do not settle how often errors will occur in a specific deployment, but they support keeping human review or other verification in workflows where incorrect answers carry costs.
As an affiliate, we earn on qualifying purchases.
How the Two Muse Versions Differ
The release follows Muse Spark 1.3 by about six weeks, according to the source report. Artificial Analysis gave the predecessor an Intelligence Index score of 48 and estimated its cost at $1.60 per index task. Muse Wonder’s score of 52 represents a higher result, while its estimated task cost is also higher. The comparison is specific to the evaluator’s methodology and should not be read as a universal measure of everyday operating costs.
The new models show different gains across reported benchmarks. Muse Wonder Full scored 57% on Terminal-Bench 4.0, up from 33% for Muse Spark 1.3, and 71% on AutomationBench-AA. The report places it third on AutomationBench-AA and AA-Briefcase, and first on GDP.pdf and CritPt. On GDPval-AA, it recorded 1,857 Elo, compared with 1,866 for Claude Opus 5.5 and 1,838 for Claude Sonnet 5.5; the report says the results are within each other’s confidence intervals.
Artificial Analysis also compared model-and-tool combinations in its Coding Agents Index. Muse Code with Muse Wonder Full scored 68, matching Claude Code with Claude Sonnet 5.5; Claude Code with Opus 5.5 scored 66. Because this index evaluates a model together with its coding harness, it does not show how Muse Wonder Full would perform inside Claude Code or how the models would compare outside those setups.
“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.
Access, Costs and Error Rates
The source material does not give a per-token or per-task price for Muse Wonder Full, nor does it say whether Meta plans to make it available through an API. It also does not provide enough detail here to independently reproduce the evaluator’s task-cost estimates or assess how those costs change across different workloads.
The report says Muse Wonder used 7.69 million input tokens per Intelligence Index task, compared with 6.03 million for Muse Spark 1.3, and that input accounted for $3.43 of its $4.52 estimated task cost. It describes the model’s output as about 54,000 tokens per task. These figures describe Artificial Analysis’s evaluation setup; the provided source excerpt ends during its cost breakdown, so further details on the calculation are not available here.
Benchmark scores and hallucination rates are also evaluator-reported measurements. The source does not establish how the models perform across all languages, industries, or production settings, and does not provide confidence intervals for every listed result. Meta’s own statements about the models’ capabilities or intended use are not included in the supplied material.
As an affiliate, we earn on qualifying purchases.
API Tests Will Set the Value
For developers, the next practical step is to compare the models on representative tasks, measuring quality, token consumption, latency and total cost rather than relying on an index score alone. That is especially relevant for Muse Wonder, whose reported benchmark cost reflects unusually high input-token use, according to the evaluation.
Further details from Meta could clarify whether Muse Wonder Full will become available beyond Muse Code and how its access will be priced. Until then, buyers can compare the published API rates for Muse Wonder with their own workload results, while treating the Full model’s benchmark standing as a result from the evaluator’s tests rather than a promise of performance in every use case.
AI performance benchmarking software
As an affiliate, we earn on qualifying purchases.
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. According to the source report, both accept text, images, audio and video, return text, and have a 1-million-token context window.
How did the models score?
Artificial Analysis gave Muse Wonder Full a score of 56 and Muse Wonder a score of 52 on its Intelligence Index v4.3. The source says Full improved on all 10 evaluations in the index.
Can customers use Muse Wonder Full through Meta’s API?
The source report says Muse Wonder Full is available only inside Muse Code. It does not say whether Meta plans to offer it through an API.
How much does Muse Wonder cost?
Its listed API rates are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimates a $4.52 cost for one task in its Intelligence Index evaluation; that is not a universal price for other tasks.
Does the evaluation show Muse Wonder is always more expensive than rivals?
No. It reports that Muse Wonder’s estimated cost was higher for the evaluator’s Intelligence Index workload than several models with similar scores. Costs can differ with token use and the work being performed.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
