🔍 Read the full analysis: Muse Wonder Full Joins The AI Frontier And Brings Muse Wonder Along on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- 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 one-million-token context window. Artificial Analysis rated Muse Wonder Full at 56 on its Intelligence Index, while the API-accessible Muse Wonder scored 52; the evaluator’s estimated cost per index task for Muse Wonder was $4.52, 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 accept text, images, audio and video. In an evaluation published October 12 by Artificial Analysis, Wonder Full scored 56 on the Intelligence Index, while the API-accessible Wonder scored 52; the evaluator estimated a Wonder task cost of $4.52, compared with $0.72 for GPT-6.1 Sol at the same score.
Both models have a one-million-token context window and return text. Meta offers Muse Wonder through its API and inside Muse Code, its coding agent. The API list price is $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Muse Wonder Full is available only to Muse Code users and has no public per-token price, so its cost per task cannot be calculated from the information provided.
Artificial Analysis reported that Wonder Full improved on all ten evaluations in its Intelligence Index v4.3. It scored 57% on Terminal-Bench 4.0, up from 33% for Muse Spark 1.3, and placed among the top three on six of the ten evaluations. It also recorded 1,857 Elo on GDPval-AA, a benchmark for economically useful knowledge work; the source says that result is within the confidence intervals of Claude Opus 5.5 and Claude Sonnet 5.5.
The evaluator’s Coding Agents Index gave Muse Code with Wonder Full 68, matching its score for Claude Code with Claude Sonnet 5.5. The index measures a model working with its software harness, rather than the model alone. Muse Code with the API model Wonder scored 63, compared with 54 for Muse Code with Muse Spark 1.3.
Muse Wonder’s Intelligence Index score of 52 matched GPT-6.1 Sol’s score in the comparison, but Artificial Analysis estimated a much higher cost per task: $4.52 for Wonder versus $0.72 for Sol. The source also reported an estimated $1.99 for Gemini 4 Argon, which scored 53. These are evaluator cost estimates for its benchmark tasks, not a universal bill for every use case.
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’s latest models into direct comparison with systems from Anthropic, Google and OpenAI. Wonder Full’s score of 56 places it near the top of the cited Intelligence Index results, while its coding-agent score suggests that the model-and-tool combination in Muse Code performs competitively on that evaluator’s measure.
Access and price complicate that picture. The higher-scoring Wonder Full is restricted to Muse Code, with no public per-token price. Developers who want API access can use Wonder, which scored four points below Full in the index and carries a higher listed token price than several models in the comparison. Artificial Analysis’ estimated task-cost figures make that difference concrete, though costs will vary with workload and usage.
The evaluation also points to a trade-off between capability and reliability. Artificial Analysis reported a 23% hallucination rate for Wonder Full and 27% for Wonder, down from 33% for Muse Spark 1.3. The improvement is meaningful in the evaluator’s comparison, but the reported rates still support retaining human review or other checks in workflows where incorrect outputs have consequences.
As an affiliate, we earn on qualifying purchases.
Six Weeks After Muse Spark
The two models arrive six weeks after Muse Spark 1.3. In Artificial Analysis’ figures, Wonder Full rose from Spark 1.3’s Intelligence Index score of 48 to 56, while API model Wonder scored 52. The source reports gains on individual measures as well, including Wonder Full’s Terminal-Bench result and Wonder’s AutomationBench-AA result of 68%, up from 58% for Spark 1.3.
Meta’s access model differs between the releases. Wonder is sold through an API at published token rates, whereas Wonder Full is offered through Muse Code without a public per-token rate. That means independent users can compare Wonder’s listed price and benchmark performance more directly than they can assess the cost of running Full.
Artificial Analysis’ comparison also includes estimated token use. It says Wonder consumed about 7.69 million input tokens per index task and about 54,000 output tokens. The source attributes $3.43, or 76%, of the estimated $4.52 task cost to input. These figures describe the evaluator’s benchmark setup; they do not establish the cost of other prompts, applications or operating conditions.
““Meta closed the capability gap. It did not close the price gap.””
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Public Pricing and Real-World Costs
Muse Wonder Full’s per-token price is not public, and the supplied material gives no cost-per-task estimate for it. That leaves users unable to make a direct price comparison between Full and other models from the figures cited here.
The benchmark results are attributed to Artificial Analysis, but the source material does not provide complete testing methods, uncertainty ranges for every score, or results across independent real-world deployments. It also does not establish how performance or costs would change across different prompts, tools and usage volumes. The task-cost figures should be read as estimates tied to the evaluator’s index, not as fixed prices for general use.
Hallucination rates and benchmark rankings do not by themselves establish how safe or reliable a model will be in a particular application. The reported figures show improvement over Muse Spark 1.3, but the source does not describe the consequences of errors in specific production workflows.
As an affiliate, we earn on qualifying purchases.
API Rollout and Further Evaluation
For now, buyers can access Muse Wonder through Meta’s API or use it in Muse Code; Wonder Full remains limited to Muse Code users. Meta has not provided a public per-token price for Full in the supplied information. A published rate or task-cost estimate would let developers compare the stronger model’s economics with competing systems.
Further independent testing can show whether the benchmark gains hold across other evaluations and practical tasks. Users weighing Wonder for production work will also need to measure their own input-token use, output volume, accuracy and review requirements, since the evaluator’s cost-per-task result may not match their workloads.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Meta release?
Meta released Muse Wonder and Muse Wonder Full. Both are proprietary models with a one-million-token context window and support text, image, audio and video input, according to the source material.
How did the models score?
Artificial Analysis reported Intelligence Index v4.3 scores of 56 for Wonder Full and 52 for Wonder. The source says Wonder Full improved on all ten evaluations in the index.
Where can users access Wonder Full?
The supplied information says Wonder Full is available only through Muse Code. It does not have a public per-token price in the source material.
Why is Muse Wonder’s cost being compared with other models?
Artificial Analysis estimated a cost of $4.52 per Intelligence Index task for Wonder, compared with $0.72 for GPT-6.1 Sol, which received the same score of 52 in the cited comparison. These are evaluator-specific estimates, not guaranteed costs for other workloads.
Are the new models more reliable than Muse Spark 1.3?
Artificial Analysis reported hallucination rates of 23% for Wonder Full and 27% for Wonder, compared with 33% for Muse Spark 1.3. Those benchmark figures indicate improvement in that evaluation, but do not guarantee reliability in every application.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
