🔍 Read the full analysis: Why Muse Wonder Full’s Arrival On The AI Frontier Matters 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
Artificial Analysis rated Meta’s Muse Wonder Full at 56 on its Intelligence Index, tying Claude Sonnet 5.5 and trailing Claude Opus 5.5 by two points. The purchasable Muse Wonder scored 52, while one benchmark task cost $4.52—far more than some models with comparable scores. Wonder Full is available only through Muse Code, and its per-task cost is not disclosed.
Meta’s new Muse Wonder Full scored 56 on Artificial Analysis’ Intelligence Index v4.3, matching Claude Sonnet 5.5 and trailing Claude Opus 5.5 by two points, according to an evaluation published October 12. The result puts Meta’s strongest model near the top of the assessed group, but the more widely available Muse Wonder API model scored 52 and was estimated to cost $4.52 per index task—a higher bill than several models with similar or better scores.
Meta shipped Muse Wonder and Muse Wonder Full six weeks after Muse Spark 1.3. Both are proprietary models with a one-million-token context window; they accept text, images, audio and video, and return text. Their availability differs: Muse Wonder can be accessed through Meta’s API and Muse Code, while Wonder Full is available only to Muse Code users.
Artificial Analysis gave Wonder Full an Intelligence Index score of 56, up from 48 for Muse Spark 1.3. The evaluator reported improvements across all ten index evaluations and placed Wonder Full in the top three on six. Its results included a 57% score on Terminal-Bench 4.0, compared with 33% for Spark 1.3, and 1,857 Elo on GDPval-AA v2.1, a measure of economically useful knowledge work. The source report says the GDPval-AA results for Wonder Full, Claude Opus 5.5 and Claude Sonnet 5.5 fell within one another’s confidence intervals.
The API model, Muse Wonder, scored 52, the same as GPT-6.1 Sol in the cited comparison. Meta lists its API 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 using Muse Wonder, compared with $0.72 for GPT-6.1 Sol, $1.99 for Gemini 4 Argon and $5.46 for Claude Sonnet 5.5. These are evaluator task-cost estimates, not a universal charge for every user request.
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
Frontier Scores Meet API Costs
The results suggest Meta has made a substantial performance advance, particularly in coding and agent-style tasks. Artificial Analysis reported that Wonder Full’s Terminal-Bench score rose by 24 percentage points from Spark 1.3. In the separate Coding Agents Index, Muse Code with Wonder Full scored 68, equal to Claude Code with Sonnet 5.5 in that evaluation. The index measures the combined model-and-software configuration, not the model in isolation, so it does not establish how Wonder Full would perform inside a different coding tool.
For customers choosing an API, the cost comparison complicates the performance story. Muse Wonder’s score of 52 matched Sol’s in the cited index, yet its estimated task cost was about 6.3 times Sol’s. The source report attributes much of Wonder’s task bill to input consumption: it processed 7.69 million input tokens per index task, with an estimated $3.43 of the $4.52 total coming from input. That is a benchmark-specific result; actual bills will depend on workload, token use and pricing conditions.
Quality also remains a consideration. Artificial Analysis reported a 23% hallucination rate for Wonder Full and 27% for Wonder, compared with 33% for Spark 1.3. Those figures indicate improvement in the evaluator’s test, not that the models are reliably correct in every setting. The source comparison also placed Wonder Full behind several named competitors on its AA-Omniscience measure. For users, the release presents a trade-off among capability, availability, cost and the need to verify outputs.
As an affiliate, we earn on qualifying purchases.
Two Models, Different Access
The release follows Muse Spark 1.3 by six weeks, according to the source report. Artificial Analysis’ evaluation compares model performance using its Intelligence Index v4.3 and estimates the cost of completing one index task using measured token use and listed prices. That framework helps compare performance and cost under a common test, but it does not predict the exact expense or results of every real-world application.
The distinction between the two new models matters for buyers. Wonder is the API product, with published per-token prices; Wonder Full is restricted to Muse Code and has no per-token price in the material provided. That means outside users cannot directly calculate Wonder Full’s cost per task from a public API price. The source report also notes that Muse Wonder’s listed input and output prices match Claude Opus 5.5, while its benchmark score is lower than Wonder Full’s and below Opus 5.5’s score of 58.
“Meta closed the capability gap. It did not close the price gap.”
— Thorsten Meyer, ThorstenMeyerAI.com
AI model performance evaluation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Costs and Real-World Performance
The supplied material does not provide an independently published per-token price or per-task cost for Muse Wonder Full, which is restricted to Muse Code. It also does not establish whether its advantage in the Coding Agents Index would hold in other coding environments or user workloads. The benchmark results measure performance on specified evaluations, not broad performance across every task.
It is also unclear how representative the reported token consumption is of typical API use. The $4.52 estimate is tied to Artificial Analysis’ Intelligence Index task and the stated list prices. Users’ costs may differ with prompt size, cached input, output length and the work being performed. The reported hallucination rates likewise do not establish the error rate for a particular deployment.
As an affiliate, we earn on qualifying purchases.
Watch Access and Pricing
The immediate next step for prospective users is to compare the models on their own workloads, including both output quality and token consumption. For API buyers, Muse Wonder’s published prices make direct cost testing possible. Muse Code users can test Wonder Full, but the supplied report gives no public per-token price with which to calculate a like-for-like bill.
Further independent evaluations and any changes to Meta’s access or pricing terms may clarify whether the benchmark gains translate into practical value. Until then, the available evidence supports a strong evaluation result for Wonder Full and a high estimated task cost for Wonder, while leaving the cost and broader performance of the Code-only model unresolved.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did Artificial Analysis find about Muse Wonder Full?
It scored 56 on the Intelligence Index v4.3, tying Claude Sonnet 5.5 in the cited comparison and finishing two points below Claude Opus 5.5.
Can customers buy Muse Wonder Full through Meta’s API?
Not according to the supplied report. Wonder Full is available only through Muse Code; Muse Wonder is the model offered through Meta’s API and Muse Code.
How much does a Muse Wonder benchmark task cost?
Artificial Analysis estimated $4.52 for one Intelligence Index task. That is a benchmark-specific cost estimate based on token use and listed prices, not a fixed fee for every request.
Why is Muse Wonder’s estimated task cost relatively high?
The report says Muse Wonder used 7.69 million input tokens per index task. It estimates that input accounted for $3.43 of the $4.52 task cost, or 76%.
Do the scores prove Muse models are more reliable?
No. The evaluation reported lower hallucination rates than Muse Spark 1.3, but still measured rates of 23% for Wonder Full and 27% for Wonder. Benchmark results do not guarantee accuracy on every real-world task.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
