🔍 Read the full analysis: Muse Wonder Full Enters AI’s Frontier: What’s Behind The Move? on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
Meta has released Muse Wonder and Muse Wonder Full, two proprietary models that accept text, images, audio and video. Artificial Analysis scored Muse Wonder Full at 56 on its Intelligence Index, but it is limited to Meta’s Muse Code, while the purchasable API model scored 52 and cost $4.52 per benchmark task in the evaluator’s analysis.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that Artificial Analysis evaluated on October 12, 2026. The evaluator gave Wonder Full an Intelligence Index score of 56, while the API-available Wonder scored 52; its analysis put Wonder’s cost at $4.52 per benchmark task, drawing attention to a gap between the model’s score and its measured price.
Artificial Analysis said Wonder Full scored eight points above Muse Spark 1.3, which received 48, and improved on all ten evaluations in the index. Among the 27 configurations included in the source comparison, only Claude Opus 5.5 at its maximum setting scored higher, with 58; Claude Sonnet 5.5 at maximum tied Wonder Full at 56. These are evaluator results, not guarantees of performance on every task.
The models differ in availability. Muse Wonder is available through Meta’s API and inside the company’s Muse Code coding agent. Its listed price is $4 per million input tokens and $20 per million output tokens, with cached input at $0.20 per million tokens. Muse Wonder Full is available only to Muse Code users, and Meta has not provided a per-token price in the source material.
Artificial Analysis also reported a score of 68 for Muse Code paired with Wonder Full in its Coding Agents Index, matching its Claude Code and Sonnet 5.5 configuration. The index evaluates the model together with its software harness, so the result does not establish how Wonder Full would perform inside another coding system. Wonder, the API model, scored 63 in that index.
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 Strength Meets API Cost
The release matters because it puts Meta’s models into a stronger position in independent evaluations while leaving users with different choices depending on access. Wonder Full’s leading result is available only through Muse Code; developers seeking API access can buy Wonder, which scored four points lower on the Intelligence Index.
Artificial Analysis estimated that one Intelligence Index task on Wonder cost $4.52, compared with $0.72 for GPT-6.1 Sol at the same score of 52, $1.99 for Gemini 4 Argon at 53, and $5.46 for Sonnet 5.5 at 56. These figures describe the evaluator’s benchmark workload, not a fixed charge for every real-world use. Actual bills vary with task length and token use.
The cost breakdown in the source points to unusually high input consumption: Wonder used 7.69 million input tokens per index task, according to Artificial Analysis. The evaluator attributed $3.43, or 76% of the $4.52 estimate, to input. That makes token efficiency and workload fit important considerations for buyers, alongside benchmark scores.
As an affiliate, we earn on qualifying purchases.
How the Two Models Compare
The release follows Muse Spark 1.3 by six weeks. In Artificial Analysis’s comparison, Wonder Full scored 57% on Terminal-Bench 4.0, up from Spark 1.3’s 33%. The evaluator reported that Wonder Full also placed among the top three on six of the ten index evaluations. On GDPval-AA, a test of economically useful knowledge work, it scored 1,857 Elo, 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.
Wonder also improved on its predecessor in several reported measures. It scored 52% on Terminal-Bench and 68% on AutomationBench-AA, compared with 58% for Spark 1.3 on the latter. Artificial Analysis reported an AA-Omniscience score of 25 for Wonder and 30 for Wonder Full. Its reported hallucination rates were 27% and 23%, respectively, down from 33% for Spark 1.3. The source cautions that those rates remain relevant for workflows where incorrect answers could carry costs.
Both models have a one-million-token context window and accept text, images, speech or audio, and video as input; both return text. Wonder’s list price is the same as the stated list price for Claude Opus 5.5. Artificial Analysis’s cost-per-task comparison is separate from those per-token prices and reflects the tokens its test tasks consumed.
As an affiliate, we earn on qualifying purchases.
Pricing and Real-World Performance
Meta’s per-token price for Muse Wonder Full is not stated, so its operating cost cannot be compared with API models using the source’s method. The material also does not specify whether Meta plans to make Wonder Full available through its API or to disclose a separate usage price.
Benchmark scores offer a controlled comparison, but they do not settle how the models will perform across customers’ varied workloads. The $4.52 task estimate depends on the test’s token consumption, and the source does not establish that every API request will use the same number of tokens. Reported hallucination rates also do not predict the risk for any one application; the figures support retaining checks for consequential outputs.
As an affiliate, we earn on qualifying purchases.
Access, Costs and Further Testing
The next practical tests will be whether Meta expands Wonder Full beyond Muse Code, publishes its pricing, or provides more detail on how developers can access it. For now, API users can assess Muse Wonder against alternatives using its published token rates and their own task patterns rather than treating the evaluator’s benchmark bill as a universal cost.
Further independent evaluations and customer experience will help show whether the reported gains hold across coding and knowledge-work tasks. Until then, the confirmed picture is split: Meta has a high-scoring model-and-harness combination in Muse Code, while its API model has a lower benchmark score and a comparatively high estimated cost in Artificial Analysis’s test.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Meta release?
Meta released Muse Wonder and Muse Wonder Full, proprietary models that accept text, images, audio and video as input and return text.
How did Muse Wonder Full score?
Artificial Analysis gave Muse Wonder Full a score of 56 on its Intelligence Index v4.3. The evaluator also reported a score of 68 for Muse Code paired with Wonder Full in its Coding Agents Index.
Can developers access Muse Wonder Full through an API?
The source says Muse Wonder Full is available only to Muse Code users. Muse Wonder is the model available through Meta’s API.
How much does Muse Wonder cost?
Meta’s listed rates are $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 task in its benchmark, a workload-specific figure rather than a flat task fee.
What remains unknown about the release?
The source does not give a per-token price for Wonder Full or say whether Meta plans to offer it through an API. It also remains to be seen how the benchmark results translate to customers’ real-world tasks and token usage.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
