🔍 Read the full analysis: Muse Wonder Full Enters AI’s Frontier In A Move That Charges on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- 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 evaluated by Artificial Analysis on October 12, 2026. Wonder Full scored 56 on the Intelligence Index, while API-available Wonder scored 52; the reported $4.52 cost per index task for Wonder is far above several models with similar scores. The evaluation is independent testing, not a guarantee of results in every use case.
Both models are proprietary and accept text, images, audio and video as input, returning text. Each has a one-million-token context window, according to the source report. Their availability differs: Muse Wonder can be accessed through Meta’s API and in Muse Code, while Wonder Full is offered only to Muse Code users.
Artificial Analysis gave Muse Wonder Full a score of 56, eight points above Muse Spark 1.3’s 48. The source report says it improved across all ten evaluations in the index and placed in the top three on six. Claude Opus 5.5 in its max configuration scored 58; Claude Sonnet 5.5 in max configuration also scored 56. These are benchmark comparisons, not direct evidence that performance will match across every product or task.
The API model, Muse Wonder, scored 52. The reported cost estimate for one Intelligence Index task was $4.52, compared with $0.72 for GPT-6.1 Sol at the same score and $1.99 for Gemini 4 Argon at 53. Artificial Analysis also reported that Muse Wonder Full used in Muse Code scored 68 on its Coding Agents Index, tying Claude Code with Sonnet 5.5. That index evaluates model-and-software combinations, not the models alone.
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 Pricing
The results put Meta’s latest models into a higher tier of independent benchmark performance than its previous release, while raising a practical question for developers: what does that performance cost to use? Wonder Full’s score of 56 is close to leading models in the report, but its restricted availability means customers cannot purchase it through a public per-token API, based on the source material.
For developers comparing services, the more immediately actionable figure is Wonder’s estimated $4.52 per benchmark task. Artificial Analysis reported $0.72 for GPT-6.1 Sol at the same index score, though a benchmark task is not necessarily representative of a customer’s workload. Pricing and cost can vary with prompt size, output length, caching and use patterns. The source’s comparison nevertheless highlights that a strong score does not by itself make an API model economical.
The reported token use offers one possible explanation. Wonder processed 7.69 million input tokens per task in the benchmark, compared with 6.03 million for Spark 1.3. Artificial Analysis attributed $3.43, or 76% of Wonder’s estimated task cost, to input. The number is tied to the evaluator’s task setup; it should not be read as a standard bill for every API request.
As an affiliate, we earn on qualifying purchases.
Two Models, Different Access
The release follows Muse Spark 1.3 by about six weeks, according to the source report. Artificial Analysis evaluated the new models using its Intelligence Index v4.3 and published its results on October 12, 2026. The report compares 27 model configurations, with Wonder Full scoring below only Claude Opus 5.5 in the cited comparison and level with Claude Sonnet 5.5 in max configuration.
Meta lists Muse Wonder at $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens, as reported by the source. The report compares that list price with Claude Opus 5.5 and says several other named competitors charge $2 per million input tokens and $10 per million output tokens. Wonder Full has no per-token price in the material and is limited to Muse Code, so its cost per task cannot be calculated from public API rates.
Artificial Analysis also reported gains in selected task results. Wonder Full scored 57% on Terminal-Bench 4.0, compared with 33% for Spark 1.3, and 71% on AutomationBench-AA. The evaluator reported an AA-Omniscience score of 30 and a hallucination rate of 23% for Wonder Full, down from 33% for Spark 1.3. These figures describe the named benchmarks and do not establish error rates in other settings.
As an affiliate, we earn on qualifying purchases.
Costs and Real-World Performance
The source material does not include direct statements from Meta or independent details about how the company trained the models. It also does not provide a public per-token price for Muse Wonder Full, so its cost relative to Wonder or competing coding agents remains unknown. The reported benchmark prices are estimates under Artificial Analysis’ evaluation method, not universal costs for a user’s application.
It is also unclear how the models perform across customers’ own workloads, how availability may change, or whether Meta will release further technical documentation. The source report gives benchmark scores and token counts, but the supplied material does not include the full methodology or confidence intervals for every result. Its comparisons should be understood within that scope.
The reported hallucination-rate improvement is encouraging but does not remove the need to verify model output. Artificial Analysis reported 23% for Wonder Full and 27% for Wonder, while the report cited 15% for Gemini 4 Argon among models scoring at least 45 on the index. Those figures depend on the evaluator’s test and should not be treated as the likelihood of an error in any particular deployment.
AI input text video audio analysis software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Availability and Pricing to Watch
The next practical developments to watch are whether Meta expands access to Muse Wonder Full, publishes a price or broader API availability, and provides additional detail about the evaluation and model capabilities. The supplied source does not state a schedule for any of those steps.
For now, developers can compare the published API rates for Muse Wonder with their own expected token volumes and test it on representative tasks. The benchmark suggests that input-token consumption may be a major cost driver for this model under Artificial Analysis’ test conditions. Any decision based on that comparison should also account for output quality, latency, reliability and the cost of verification in the intended application.
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, according to the source report. Wonder is available through Meta’s API and Muse Code; Wonder Full is available only in Muse Code.
How did Muse Wonder Full score?
Artificial Analysis reported a score of 56 on its Intelligence Index v4.3, eight points above Muse Spark 1.3. Claude Opus 5.5 in max configuration scored 58, and Claude Sonnet 5.5 in max configuration also scored 56.
How much does Muse Wonder cost?
The source reports list prices of $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimated $4.52 for one Intelligence Index task, a figure tied to its test workload.
Can customers buy Muse Wonder Full through an API?
The supplied report says Muse Wonder Full is available only to Muse Code users and has no per-token price. It does not provide information about a planned public API release.
Do the benchmark results show how the models will perform for every user?
No. The figures are results from specific Artificial Analysis evaluations. Performance, cost and error rates can differ across tasks and real-world deployments, so the results do not guarantee outcomes for an individual application.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
