Muse Wonder Full Arrives In AI As Muse Wonder Charges Up

🔍 Read the full analysis: Muse Wonder Full Arrives In AI As Muse Wonder Charges Up on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

Meta has released Muse Wonder and Muse Wonder Full, two proprietary AI models that an Artificial Analysis evaluation published October 12 rated at 52 and 56, respectively, on its Intelligence Index. The stronger model is available only through Meta’s Muse Code product, while the API-accessible Wonder was estimated to cost $4.52 per benchmark task—far more than GPT-6.1 Sol, which scored the same 52 in the comparison.

Artificial Analysis said Muse Wonder Full improved on all ten evaluations in its Intelligence Index v4.3. Its score of 56 was eight points above Muse Spark 1.3’s 48. In the source’s comparison of 27 model configurations, Claude Opus 5.5 at its maximum setting scored 58, while Claude Sonnet 5.5 at its maximum setting also scored 56. The scores are benchmark results, not a guarantee of performance on every task.

The models differ in access and pricing. Muse Wonder is available through Meta’s API and inside Muse Code, with listed prices of $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens. Muse Wonder Full is limited to Muse Code and has no per-token price disclosed in the source, so an equivalent API cost cannot be calculated.

Both models have a one-million-token context window and accept text, images, speech or audio, and video as input; they return text. Artificial Analysis estimated that one Intelligence Index task on Muse Wonder cost $4.52, against $0.72 for GPT-6.1 Sol, $1.99 for Gemini 4 Argon and $5.98 for Claude Opus 5.5. These figures apply to the evaluator’s benchmark workload and should not be read as fixed costs for ordinary use.

Performance Comes With a Cost

The release gives Meta a stronger showing in independent model evaluations, but it also highlights a gap between the performance of the model most users can access and the one Meta keeps inside its coding product. Muse Wonder Full’s 56 puts it alongside high-scoring configurations from Anthropic on this index; the API model’s 52 matches GPT-6.1 Sol’s score in the cited comparison.

For customers deciding which API to use, the reported task-cost comparison is material: Artificial Analysis estimated $4.52 for Muse Wonder versus $0.72 for Sol on its Intelligence Index workload. The source also says Wonder used 7.69 million input tokens per task, more than the compared Sol, Astra and Argon models. That token use may help explain the benchmark bill, though the supplied material does not establish how the estimate would translate to a particular customer’s requests.

Performance figures also do not remove reliability concerns. The evaluation put Muse Wonder’s hallucination rate at 27% and Wonder Full’s at 23%, down from 33% for Muse Spark 1.3. The source’s figures still indicate that confident errors remain possible. Users applying these systems to consequential work would need to check outputs rather than assume that a higher benchmark score guarantees correctness.

From Spark to Two Models

The release follows Muse Spark 1.3 by about six weeks, according to the source article. The reported Intelligence Index score rose from 48 for Spark 1.3 to 52 for Muse Wonder and 56 for Wonder Full. The two new models therefore represent different points in Meta’s product offering: an API model with published token prices and a higher-scoring model available through Muse Code.

Artificial Analysis’ evaluation includes multiple task-specific measures, not just an overall index score. The source reports that Wonder Full scored 57% on Terminal-Bench 4.0, up from 33% for Spark 1.3, and finished in the top three on six of ten index evaluations. On the Coding Agents Index, which evaluates a model together with its software harness, Muse Code paired with Wonder Full scored 68, tying the cited Claude Code and Sonnet 5.5 configuration. That result does not establish how Wonder Full would perform inside Claude Code.

The source also reports that Wonder Full scored 30 on AA-Omniscience, compared with 25 for Muse Wonder, while listing higher scores for several rival models. These task-specific results add detail to the overall ratings but do not by themselves establish a universal ranking across real-world uses.

“That gap is the story.”

— Thorsten Meyer, ThorstenMeyerAI.com

Pricing and Reliability Questions

The supplied material does not disclose a per-token or per-task price for Muse Wonder Full, which is available only inside Muse Code. It is therefore not possible to compare its operating cost directly with the API models using the figures provided. Meta’s own explanation for Wonder’s pricing, including whether the rates reflect expected use, infrastructure costs or product positioning, is also not included.

The $4.52 task estimate is specific to Artificial Analysis’ Intelligence Index workload. Actual costs depend on the requests made, the number of tokens processed and the applicable pricing. The source begins a cost breakdown—reporting $3.43, or 76%, for input—but its supplied text ends before the remaining breakdown is complete.

Benchmark scores and hallucination rates also have limits. The source does not provide full uncertainty ranges for every result or details sufficient to generalize the reported error rates to all use cases. It remains unclear how the models perform across particular customer workloads, how availability may change, and whether Meta will offer Wonder Full outside Muse Code.

Watch for Access and Cost Details

The immediate next step for prospective users is to compare the published API rates with their own expected token use and workload, rather than treating the benchmark task estimate as a standard charge. Customers can also test Muse Wonder against alternatives on the tasks they actually need, with independent verification for outputs where errors carry a cost.

Further details from Meta could clarify whether Wonder Full will remain exclusive to Muse Code, whether its pricing will be disclosed, and how the company arrived at Muse Wonder’s API rates. Additional independent evaluations would help show whether the benchmark gains hold across other tests and real-world applications. No expansion of access or pricing change is confirmed in the supplied material.

Key Questions

What are Muse Wonder and Muse Wonder Full?

They are two proprietary Meta AI models released after Muse Spark 1.3. Both accept text, images, audio or speech, and video as input, return text, and have a one-million-token context window.

Which model can customers access through Meta’s API?

Muse Wonder is available through Meta’s API and inside Muse Code. Muse Wonder Full is available only through Muse Code, according to the source.

How did the models score in the reported evaluation?

Artificial Analysis gave Muse Wonder Full a score of 56 and Muse Wonder a score of 52 on its Intelligence Index v4.3. Those are benchmark results and do not establish performance on every task.

How much does Muse Wonder cost compared with GPT-6.1 Sol?

Artificial Analysis estimated a cost of $4.52 per Intelligence Index task for Muse Wonder and $0.72 for GPT-6.1 Sol, with both scoring 52 in the cited comparison. The estimate reflects the benchmark workload, not a fixed price for all customer uses.

Is Muse Wonder Full’s price known?

No per-token price is provided in the source. Because Wonder Full is described as available only in Muse Code, the supplied information does not allow a direct cost comparison with API models.

Source: ThorstenMeyerAI.com