🔍 Read the full analysis: Muse Wonder Full Joins AI’s Frontier In A Charged New Move on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Artificial Analysis gave Meta’s Muse Wonder Full a score of 56 on its Intelligence Index, placing it near leading Claude models. The API-available Muse Wonder scored 52, but an evaluated task cost $4.52—far more than similarly scoring GPT-6.1 Sol. Muse Wonder Full is restricted to Muse Code, and its per-task cost is not publicly available.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that an evaluation published by Artificial Analysis on October 12 places near leading commercial systems. Muse Wonder Full scored 56 on the Intelligence Index, while the model customers can access through Meta’s API scored 52; the evaluated API task cost was $4.52, substantially more than a similarly scoring competitor.
Artificial Analysis says Muse Wonder Full improved on all 10 evaluations in its Intelligence Index v4.3. It ranked in the top three on six, including Terminal-Bench 4.0, where it scored 57%, up from 33% for Muse Spark 1.3. The evaluation lists Claude Opus 5.5 at 58 and Claude Sonnet 5.5 at 56, making Wonder Full’s score close to those systems. These are benchmark results, not a guarantee of performance in every use.
The two models differ in access. Muse Wonder is available through Meta’s API and Muse Code, while Muse Wonder Full is available only through Muse Code. Both have a reported one-million-token context window and accept text, images, audio and video as input, returning text. Meta’s listed Wonder API rates are $4 per million input tokens, $0.20 for cached input and $20 per million output tokens.
Artificial Analysis estimates that one Intelligence Index task costs $4.52 on Muse Wonder. GPT-6.1 Sol scored the same 52 for an estimated $0.72 per task; Gemini 4 Argon scored 53 for $1.99. Muse Wonder Full has no per-token price listed because it is restricted to Muse Code, so its cost per benchmark task cannot be compared from the supplied figures.
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
The release suggests Meta has produced a model that can compete on several measured tasks, but the strongest results belong to the version that is not sold through a public API. Muse Wonder Full’s score of 56 and its Coding Agents Index score of 68, tied with the listed Claude Code and Claude Sonnet 5.5 configuration, give Muse Code users a competitive option. The coding index measures a model working with its software harness, so it does not establish how Wonder Full would perform in another coding environment.
For developers choosing an API, the distinction is practical: Muse Wonder scored 52, and the source’s task-cost estimate is higher than those for several models with similar or better index scores. Artificial Analysis attributes much of the bill to input volume: Wonder processed 7.69 million input tokens per task, accounting for a reported $3.43, or 76% of the $4.52 estimate. That makes the cost comparison relevant to teams weighing benchmark performance against usage expense.
The results also leave reliability as a consideration. Artificial Analysis reports a 27% hallucination rate for Wonder and 23% for Wonder Full, down from 33% for Spark 1.3. The improvement is a benchmark finding, not proof that generated answers are dependable in every setting. Workflows where errors carry financial or operational consequences still need suitable checks.
As an affiliate, we earn on qualifying purchases.
Six Weeks After Spark 1.3
The two models arrived six weeks after Muse Spark 1.3, according to the source report. Artificial Analysis’ comparison says Wonder Full’s Intelligence Index score rose from 48 for Spark 1.3 to 56. The same evaluation reports a Terminal-Bench increase from 33% to 57% and a reduction in hallucination rate from 33% to 23%. These figures describe performance under the named evaluation methods and should not be treated as universal measures of everyday use.
Artificial Analysis reports a score of 52 for Muse Wonder and 56 for Wonder Full. Its cost comparison estimates $4.52 for a Wonder task, against $1.60 for Spark 1.3. The report also gives Wonder’s list prices as $4 per million input tokens and $20 per million output tokens, compared with Spark 1.3’s $1.25 and $4.25. It says the Wonder rates match Claude Opus 5.5’s listed rates; the per-task estimates depend on the models’ token use as well as their prices.
Wonder Full’s evaluation was particularly strong on some tasks: Artificial Analysis reports 33% on GDP.pdf and 33% on CritPt, ranking it first in both comparisons described in the source. On GDPval-AA, which evaluates economically useful knowledge work, it scored 1,857 Elo, between Claude Opus 5.5 at 1,866 and Sonnet 5.5 at 1,838. The source notes those results fall within the models’ confidence intervals.
“The gap is the story.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Costs and Real-World Reliability
The supplied report does not provide Meta’s own benchmark methodology or a response from the company about the model release, pricing rationale or evaluation results. The reported scores and task costs are attributed to Artificial Analysis; the cost figures are estimates based on benchmark runs and token consumption, not a promise that every customer task will cost the same amount.
The commercial cost of Muse Wonder Full is also not disclosed on a per-token basis. Since access is limited to Muse Code, readers cannot use the supplied figures to compare its per-task price directly with API models. It is not clear from the material whether Meta plans to expand access or publish additional pricing information.
Benchmark results do not settle how the models will perform across customer workloads. The reported hallucination rates remain nonzero, and the available material does not specify how results vary by task, prompt or deployment conditions. The extent to which the benchmark gains translate into better outcomes for users remains to be established.
As an affiliate, we earn on qualifying purchases.
Access and Pricing to Watch
The immediate next step for customers is to test Muse Wonder through Meta’s API or use Muse Wonder Full within Muse Code, while comparing costs on their own workloads rather than relying only on a benchmark estimate. Teams using either model for consequential work will need to assess output quality and maintain verification appropriate to the task.
Further clarity depends on Meta. A broader release of Wonder Full, public pricing for that model, or detailed company comments on the evaluation could change how customers judge the offering. The supplied report does not announce a date for any such development, so future access and pricing plans remain unconfirmed.
AI performance evaluation 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. Both accept text, images, audio and video as input and return text, with a reported one-million-token context window.
How do the models differ in availability?
Muse Wonder is available through Meta’s API and Muse Code. Muse Wonder Full is available only through Muse Code, according to the source report.
How did Muse Wonder Full score?
Artificial Analysis gave it 56 on the Intelligence Index v4.3. The evaluation says it improved on all 10 index tasks and ranked in the top three on six.
Why is Muse Wonder’s estimated task cost drawing attention?
Artificial Analysis estimates $4.52 per Intelligence Index task for Muse Wonder, compared with $0.72 for GPT-6.1 Sol, which scored 52 as well. These are evaluation-based estimates, and actual costs depend on usage.
Is Muse Wonder Full’s API price known?
No per-token price is provided in the source material. The model is reported as available only through Muse Code, so its API task cost cannot be compared using the figures supplied.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
