🔍 Read the full analysis: Muse Wonder Full Joins The AI Frontier: How Muse Wonder Fits In 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
Artificial Analysis reported that Meta’s Muse Wonder Full scored 56 on its Intelligence Index, while Muse Wonder, the model available through Meta’s API, scored 52. The evaluation also estimated $4.52 per Intelligence Index task for Muse Wonder, far above some similarly scoring rivals; Muse Wonder Full has no public per-token price and is available only in Muse Code.
Artificial Analysis’ Intelligence Index v4.3 gave Muse Wonder Full a score of 56, eight points above the previous Muse Spark 1.3’s 48. The source report says Wonder Full improved across all ten evaluations in the index and ranked among the top three on six. Among 27 model configurations compared by the report, only Claude Opus 5.5 in its max configuration scored higher, at 58; Claude Sonnet 5.5 max also scored 56. These are benchmark results, not a guarantee of performance in every use case.
The two models share a one-million-token context window and accept text, images, audio and video as input, returning text. Their access differs: Muse Wonder is available through Meta’s API and in Muse Code, while Wonder Full is limited to Muse Code users. The report lists Muse Wonder at $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. It says Wonder Full has no per-token price, so an outside user cannot calculate its cost per task from a public rate.
Several reported evaluation results point to gains in coding and work tasks. Wonder Full scored 57% on Terminal-Bench 4.0, compared with 33% for Muse Spark 1.3, and 1,857 Elo on GDPval-AA, a knowledge-work evaluation. Its coding-agent configuration, Muse Code plus Wonder Full, scored 68 in the Artificial Analysis Coding Agents Index, tying the listed Claude Code plus Sonnet 5.5 configuration. That index evaluates the model and its software harness together, so it does not establish how Wonder Full would perform inside a different coding agent.
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 Meet API Pricing
The release puts Meta in a stronger position on the reported benchmarks, but the model most users can purchase did not match Wonder Full’s score. Muse Wonder scored 52, the same as GPT-6.1 Sol in the source’s comparison, while its estimated cost was $4.52 per Intelligence Index task, against $0.72 for Sol. Artificial Analysis also listed Gemini 4 Argon at $1.99 per task with a score of 53. These figures describe the specific evaluation and cost methodology in the report; they are not universal costs for every prompt or workload.
For businesses comparing models, benchmark quality is only one part of the decision. Token consumption, API rates, reliability and the cost of checking outputs can affect whether a model is practical at scale. The source says Muse Wonder used 7.69 million input tokens per index task, compared with 6.03 million for Spark 1.3, and that input accounted for 76% of Wonder’s estimated task bill. That makes usage patterns and token volume relevant alongside list prices.
The results also leave users with a product-access distinction: the higher-scoring Wonder Full is not available as a separately priced API model, while the API version costs more in the report’s comparison than some alternatives with similar index scores. The evaluation offers evidence of progress, but it does not settle how the models perform across individual industries, customer workloads or production conditions.
As an affiliate, we earn on qualifying purchases.
From Spark 1.3 to Wonder
The source places the launch roughly six weeks after Muse Spark 1.3. Artificial Analysis’ comparison found broad gains for Wonder Full, including the Terminal-Bench increase from 33% to 57%. Muse Wonder also improved on selected measures over Spark 1.3: the report lists its AutomationBench-AA score at 68%, up from 58%, and its hallucination rate at 27%, compared with 33% for the earlier model.
Performance was not uniform across the field. On Artificial Analysis’ AA-Omniscience measure, which the report describes as rewarding correct answers and penalizing confident errors, Wonder Full scored 30 and Wonder scored 25. The listed scores for Claude Opus 5.5, GPT-6 Astra and Gemini 4 Argon were 46, 43 and 42. The report also gives hallucination rates of 23% for Wonder Full and 27% for Wonder, compared with 15% for Gemini 4 Argon among the models in its stated comparison group. Those rates are tied to the evaluation methodology and should not be treated as a prediction for every use.
“The bill is in the input, not the output.”
— Thorsten Meyer
As an affiliate, we earn on qualifying purchases.
Costs and Real-World Results
Muse Wonder Full’s per-token price is not public in the source material, so its cost per task cannot be directly compared with the API model or competitors. The reported $4.52 figure for Muse Wonder is an estimate for one Intelligence Index task under Artificial Analysis’ testing and pricing assumptions; the source excerpt does not provide the complete breakdown, ending mid-sentence after stating that input represented $3.43, or 76%, of the total.
The evaluation also does not establish how the models will perform on specific customer tasks, how much outputs may vary with different prompts, or whether Meta’s API pricing or access terms will change. The cited source provides no direct statement from Meta about the benchmark results, pricing rationale or plans to make Wonder Full available through the API. Users should treat the benchmark and cost figures as reported measurements rather than independent guarantees of business value.
As an affiliate, we earn on qualifying purchases.
Access, Pricing and Further Testing
The immediate next step for prospective users is to compare the models on their own workloads, including the amount of input data they send, output length, accuracy requirements and the cost of verification. Muse Wonder can be tested through Meta’s API or Muse Code, according to the source; Wonder Full access is limited to Muse Code. The report does not give a timetable for wider access or a public API price for the full model.
Further evaluation results, any change in Meta’s pricing or availability, and independent testing on production tasks would clarify whether the benchmark gains translate into practical value. Until then, the published comparison establishes strong results for Wonder Full and a higher estimated task cost for the purchasable Wonder model, while leaving the full model’s external economics unresolved.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the main news about Meta’s Muse models?
Artificial Analysis published an evaluation on October 12, 2026, scoring Muse Wonder Full at 56 and Muse Wonder at 52 on its Intelligence Index. The source report also highlights a $4.52 estimated cost per index task for the API model.
Can users access Muse Wonder Full through Meta’s API?
Not according to the source material. It says Muse Wonder Full is available only to Muse Code users; Muse Wonder is the version available through Meta’s API and Muse Code.
How much does Muse Wonder cost?
The source lists Muse Wonder at $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; actual costs depend on usage, and the figure is specific to that evaluation.
How does Muse Wonder compare with competing models?
In the reported Intelligence Index results, Muse Wonder scored 52, matching GPT-6.1 Sol’s score, while the source estimated their task costs at $4.52 and $0.72, respectively. Muse Wonder Full scored 56, tying Claude Sonnet 5.5 max in the comparison. Benchmark scores and estimated task costs do not determine performance or expense for every real-world workload.
What remains unknown about the release?
The source does not provide a public per-token price for Wonder Full, a timetable for API access, or Meta’s response to the evaluation and pricing comparisons. The results also do not show how the models will perform across individual customer tasks.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
