🔍 Read the full analysis: Muse Wonder Full Takes Its Place On The AI Frontier 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 that Artificial Analysis evaluated on October 12, 2026. Wonder Full scored 56 on the Intelligence Index, while the API-available Wonder scored 52; the latter’s estimated benchmark-task cost was $4.52, 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 52 and 56, respectively, on the Artificial Analysis Intelligence Index v4.3. The results place Wonder Full among leading evaluated models, while the API-accessible Wonder carries an estimated $4.52 cost per benchmark task—far above GPT-6.1 Sol’s $0.72 for the same index score of 52, according to the evaluation reported by ThorstenMeyerAI.com.
The independent evaluation, published by Artificial Analysis on October 12, 2026, found that Muse Wonder Full scored eight points above its predecessor, Muse Spark 1.3, and improved across all ten evaluations in the index. Among 27 configurations compared in the source report, Claude Opus 5.5 in its max configuration scored higher at 58; Claude Sonnet 5.5 max also scored 56. Those are benchmark results, not a guarantee of performance on every task.
The models differ in access. Muse Wonder is available through Meta’s API and in Muse Code, the company’s coding agent. Meta lists it at $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Muse Wonder Full is available only to Muse Code users, with no per-token price published in the source material. Both models have a reported one-million-token context window, accept text, images, audio and video as input, and return text.
Wonder Full’s reported scores include 57% on Terminal-Bench 4.0, up from 33% for Muse Spark 1.3, and 1,857 Elo on GDPval-AA, which measures economically useful knowledge work. The report says Wonder Full placed in the top three on six of the index’s ten evaluations. In Artificial Analysis’s Coding Agents Index, Muse Code with Wonder Full scored 68, tied with Claude Code using Sonnet 5.5. That comparison evaluates model-and-harness combinations, rather than the models in identical software environments.
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
A Stronger Model, With a Cost Gap
The release puts Meta closer to established competitors on several reported benchmark measures, but the two-model lineup creates a practical distinction: the higher-scoring Wonder Full is confined to Muse Code, while the model available through the API scored lower and has a comparatively high estimated cost per index task. Businesses weighing a model for repeated workloads may care as much about cost per task as headline benchmark scores.
Artificial Analysis estimated a task cost of $4.52 for Muse Wonder, compared with $0.72 for GPT-6.1 Sol, which also scored 52. The source report puts Gemini 4 Argon at $1.99 for a score of 53, and Claude Sonnet 5.5 at $5.46 for a score of 56. These are estimates tied to the evaluation’s workload and token use, not universal prices for every application. Still, they suggest that Meta’s API list rates and the model’s input-token consumption could affect the economics of sustained use.
The reported reliability results also call for care. Muse Wonder Full’s hallucination rate was 23%, down from 33% for Muse Spark 1.3; Wonder’s was 27%. The report says Gemini 4 Argon had a 15% rate among models scoring at least 45 on the Intelligence Index. Lower hallucination rates do not remove the need to verify outputs, particularly in workflows where errors carry financial or operational consequences.
As an affiliate, we earn on qualifying purchases.
How the Two Models Differ
Meta shipped the models six weeks after Muse Spark 1.3. The source report describes both as proprietary and says each supports a one-million-token context window and text, image, speech or audio, and video inputs. Their availability is not the same: Wonder can be called through Meta’s API and used in Muse Code, whereas Wonder Full is limited to Muse Code users and has no disclosed token-based price.
The benchmark report highlights a large change in terminal-based task performance: Wonder Full scored 57% on Terminal-Bench 4.0, compared with 33% for Spark 1.3. Wonder scored 52% on that test. On Artificial Analysis’s Coding Agents Index, Muse Code paired with Wonder Full tied the reported score of Claude Code paired with Sonnet 5.5, at 68. Because that index tests the model together with its software harness, the result does not establish how Wonder Full would perform inside Claude Code or in other setups.
Cost estimates in the report reflect the number of tokens consumed on an Intelligence Index task as well as listed token rates. Muse Wonder used 7.69 million input tokens per task, compared with 6.03 million for Spark 1.3, according to the source. The report attributes $3.43 of Wonder’s $4.52 estimated task cost to input tokens. That makes its benchmark-task estimate specific to the evaluation’s workload; it should not be read as a fixed charge for any individual API request.
“The difference is the bill.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Open Questions on Price and Performance
The source material does not provide a per-token price for Muse Wonder Full, so customers cannot compare its direct API cost with the listed rates of competing models. It also does not state when, or whether, Meta will make Full available outside Muse Code. Meta’s explanation for Wonder’s higher list price and input-token use is not included.
The reported benchmark scores and task-cost estimates describe Artificial Analysis’s specific tests. The source does not provide enough detail here to establish how results would transfer to every customer’s prompts, tools, or workloads. The figures should also be understood as a dated snapshot: the evaluation was published on October 12, 2026, and model offerings and prices may change.
Performance measures do not settle questions about reliability in production. The reported hallucination rates vary across models and evaluations, and the material does not describe how they were measured in detail. It remains unclear how Wonder and Wonder Full will perform on independent customer workloads or how their error rates will compare in particular applications.
As an affiliate, we earn on qualifying purchases.
What Customers Can Test Now
Customers can assess Muse Wonder through Meta’s API or try either model in Muse Code, subject to the access described in the source material. For a purchasing decision, the most useful next step is to test representative tasks and track both output quality and token consumption. The report’s $4.52 task estimate can inform comparison, but it is not a substitute for workload-specific costs.
Further details from Meta on Wonder Full’s pricing, availability beyond Muse Code and the reasons behind Wonder’s token use would help clarify the commercial picture. Additional independent evaluations and customer results could show whether the benchmark gains carry over to everyday coding and knowledge-work tasks. Until then, the confirmed picture is a strong reported benchmark showing for Wonder Full and a higher estimated task cost for the API-accessible Wonder.
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 released six weeks after Muse Spark 1.3. Both accept text, images, audio and video as input, return text, and have a reported one-million-token context window.
Which model can customers access through Meta’s API?
Muse Wonder is available through Meta’s API and in Muse Code. The higher-scoring Muse Wonder Full is reported as available only to Muse Code users.
How did the models score?
In Artificial Analysis’s Intelligence Index v4.3, Muse Wonder Full scored 56 and Muse Wonder scored 52. The evaluation was published on October 12, 2026; results reflect that benchmark and date.
Why is Muse Wonder’s estimated task cost drawing attention?
The source report estimates $4.52 per Intelligence Index task for Muse Wonder, versus $0.72 for GPT-6.1 Sol, although both scored 52. The estimate depends on the benchmark workload and token use, so it is not a fixed cost for every API request.
Is Muse Wonder Full’s price known?
No per-token price for Muse Wonder Full is provided in the source material. It is reported as available only through Muse Code, leaving its cost per task unclear.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
