🔍 Read the full analysis: Muse Wonder Full’s Entry Into AI Comes As Muse Wonder Charges 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 reported strong benchmark results for Meta’s new Muse Wonder models on October 12, 2026. Muse Wonder Full scored 56 on its Intelligence Index but is available only through Muse Code; the API model, Muse Wonder, scored 52 and cost $4.52 per index task in the analysis, compared with $0.72 for GPT-6.1 Sol at the same score.
Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that scored strongly in an Artificial Analysis evaluation published October 12, 2026. The stronger model scored 56 on the group’s Intelligence Index but is available only through Meta’s Muse Code coding agent; the API-accessible Muse Wonder scored 52 and cost $4.52 per benchmark task in the evaluation, compared with $0.72 for GPT-6.1 Sol at the same score.
Artificial Analysis said Muse Wonder Full scored 56 on its Intelligence Index v4.3, up from 48 for Muse Spark 1.3. The source report says Full improved on all ten evaluations in the index and placed among the top three on six. It scored 57% on Terminal-Bench 4.0, compared with 33% for Spark 1.3, and achieved 1,857 Elo on GDPval-AA, a measure of economically useful knowledge work. The report places that GDPval-AA result between Claude Opus 5.5 at 1,866 and Claude Sonnet 5.5 at 1,838, within the reported confidence intervals.
The commercially accessible model scored 52 on the same index, matching GPT-6.1 Sol in the report. Meta lists Muse Wonder API pricing at $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimated a cost of $4.52 for one Intelligence Index task using Muse Wonder. Its comparison put GPT-6.1 Sol at $0.72 per task and Gemini 4 Argon at $1.99, while Claude Sonnet 5.5 scored four points higher than Muse Wonder at 56 for $5.46 per task.
Muse Wonder Full has no per-token public price in the source material and is limited to Muse Code users. In Artificial Analysis’ Coding Agents Index, Muse Code paired with Wonder Full scored 68, matching Claude Code with Claude Sonnet 5.5. The report cautions that this measures a model together with its coding-agent harness; it does not establish how Wonder Full would perform inside Claude Code.
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 and Cost
The release puts Meta’s model capability and its commercial offer on different footing. Muse Wonder Full’s score suggests a competitive coding-agent configuration, but customers cannot obtain that model through the public API or compare its per-token cost using the information provided. Buyers who need API access instead have Muse Wonder, whose benchmark score is lower and whose estimated task cost exceeded several alternatives in the report.
That difference matters to organizations choosing models for repeated work: headline benchmark performance does not show the full cost of operating a system. Artificial Analysis estimated that a Muse Wonder index task used 7.69 million input tokens, accounting for $3.43, or 76% of the reported $4.52 task cost. The report says Muse Wonder Full used 6.44 million input tokens per task. Those figures make token consumption, not just list rates or benchmark scores, central to assessing expenses.
The scores also do not remove reliability concerns. The report gives Muse Wonder a 27% hallucination rate and Wonder Full 23%, down from 33% for Spark 1.3. These are results under the report’s evaluation, not a guarantee about any particular deployment. For work where errors carry costs, users may still need independent checks.
As an affiliate, we earn on qualifying purchases.
From Spark 1.3 to Two Models
The source report says Meta released the two models six weeks after Muse Spark 1.3. Both are proprietary, have a one-million-token context window, and accept text, images, speech or audio, and video as input; both return text. Their availability differs: Muse Wonder is offered through Meta’s API and Muse Code, while Wonder Full is limited to Muse Code.
Artificial Analysis’ Intelligence Index combines ten evaluations, so its overall score is a summary rather than a result from a single task. The report also records specific strengths and limitations. Wonder Full scored 57% on Terminal-Bench 4.0 and 71% on AutomationBench-AA; Muse Wonder scored 52% and 68%, respectively. On AA-Omniscience, which the source describes as rewarding correct answers and penalizing confident wrong ones, the models scored 30 and 25. Those metrics provide context for the overall scores, but do not by themselves establish how either model will perform across every real-world task.
The report’s cost comparison uses the estimated expense of running one Intelligence Index task, not a universal price for every customer workload. The amount can vary with task design and token use. Meta’s stated API rates are separate from that benchmark estimate and should not be confused with a fixed charge for all uses.
AI model performance evaluation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Pricing and Performance Questions Remain
The source material does not give a per-token price for Muse Wonder Full, so customers cannot compare its direct API-style cost with other models from the information provided. It also does not specify whether Meta plans to make Full available outside Muse Code, or whether access and pricing will change.
The reported task-cost estimates reflect Artificial Analysis’ benchmark setup. The source does not provide enough detail here to determine how closely those costs match a customer’s own workloads, where prompts, context length, caching and output requirements may differ. Nor does a benchmark score establish the outcome for every coding or business task. The release’s longer-term adoption, usage and production reliability remain unknown.
As an affiliate, we earn on qualifying purchases.
Watch for API Access and Cost Data
The next practical developments to watch are whether Meta publishes pricing or broader access for Muse Wonder Full, and whether it provides more detail on the costs of using that model. For API customers, actual token consumption and the bill for their own tasks will matter alongside the listed rates and benchmark comparisons.
Further independent testing could show whether the benchmark results hold across additional workloads and how the models perform in routine deployments. Until then, the confirmed picture is limited to Meta’s availability and stated API prices, plus Artificial Analysis’ evaluation and task-cost estimates published on October 12, 2026.
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 and support text, image, audio or speech, and video input, according to the source report.
How did the models score?
Artificial Analysis gave Muse Wonder Full a score of 56 and Muse Wonder a score of 52 on its Intelligence Index v4.3. The report says Muse Wonder’s score matched GPT-6.1 Sol’s.
Can customers access Muse Wonder Full through Meta’s API?
The source report says Muse Wonder Full is available only through Muse Code. Muse Wonder is the model available through Meta’s API and Muse Code.
How much does Muse Wonder cost?
Meta’s listed API rates are $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Artificial Analysis estimated $4.52 to run one Intelligence Index task with Muse Wonder; that is a benchmark-specific estimate, not a fixed cost for every task.
What remains unknown about Muse Wonder Full?
The source material gives no per-token price for Wonder Full and does not say whether Meta plans to offer it beyond Muse Code. Its costs and performance on customers’ own workloads are also not established by the cited benchmark.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
