🔍 Read the full analysis: How Effort Level Shapes What You Spend On Claude Fable 5.2 on ThorstenMeyerAI.com
Get the little things that make your day 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 reports that Claude Fable 5.2’s score on its Intelligence Index ranges from 54 at Low effort to 65 at Max, while estimated cost per benchmark task rises from $3.10 to $26.54. The figures show that the most expensive setting does not automatically deliver the best result for every task; the benchmark costs are estimates, not a fixed charge for typical user requests.
Claude Fable 5.2 can cost an estimated $3.10 to $26.54 per task on Artificial Analysis’s Intelligence Index, depending on its selected effort level, according to figures published November 10. The benchmark scores rise from 54 at Low to 65 at Max, giving businesses a measure of the trade-off between model effort, performance and cost—not a guarantee of results or a standard price for every request.
Artificial Analysis tested five settings—Low, Medium, High, Xhigh and Max—across nine evaluations spanning agentic coding, business documents, science and knowledge. Its reported average cost per task is $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. The corresponding scores are 54, 56, 62, 62 and 65. The index rank for Max was first among 74 models, while High ranked third and Xhigh second, according to the supplied Artificial Analysis figures.
The cost gap reflects how much output the model uses while working through the test, not a different listed token rate for each setting. Anthropic’s stated rates for Fable 5.2 are $8 per million input tokens and $40 per million output tokens. In the full index run, Artificial Analysis recorded 34.9 million output tokens at Low and 275 million at Max. The reported average time per task also increased, from 6.5 minutes at Low to 46.4 minutes at Max.
The figures do not show a steady score gain at every step. Moving from Medium to High adds six points for an estimated $5.28 more per task. Xhigh costs more than High but has the same rounded score of 62 in the supplied table; Max costs nearly three times as much as High for three additional points. On Artificial Analysis’s FrontierCode 2.0 coding test, the source says High scored 49.1%, compared with 48.8% at Max. These results apply to the named tests and should not be treated as a universal ranking of settings for real-world work.
How Effort Shapes Claude Fable 5.2 Costs Performance has a price
Artificial Analysis estimates task costs rise with effort, while benchmark scores do not improve at every step. These averages are a comparison point—not a fixed price for typical requests.
Cost rises across effort settings
Artificial Analysis reports average API costs across nine benchmark evaluations.
Score gains are uneven
More effort can improve the index score, but higher cost does not guarantee a better result on every test.
| Setting | Score | Estimated cost |
|---|---|---|
| Low | 54 | $3.10 |
| Medium | 56 | $3.80 |
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
What businesses should weigh
The right setting depends on each workflow’s quality, cost and turnaround needs.
- Can raise benchmark scores
- Costs more per benchmark task
- Max averaged 46.4 minutes per task
- Costs less in the reported tests
- May suit routine, high-volume work
- Low averaged 6.5 minutes per task
Test on your own work
Benchmark averages are a starting point, not a prediction for every customer prompt.
Use real work patterns to compare effort settings.
Record results, completion time and API costs.
Keep a higher setting only when its added value justifies the cost.
The Cost of More Model Effort
For organizations using Claude through an API, the results put a practical choice alongside the headline performance score: how much additional reasoning is worth paying for? A workflow that handles routine requests in volume may be sensitive to per-task costs, while a smaller number of difficult tasks could justify more computation if it improves results. The benchmark offers a way to compare settings, but it cannot by itself establish the right choice for a company’s own workload.
The time figures matter as well as the estimated bill. A task taking nearly 46 minutes on average at Max could be a poor fit for work that depends on quick responses, even if its benchmark score is higher. Conversely, a slower setting might be acceptable for a task where quality matters more than turnaround. Those are operational judgments, not conclusions proven by the index.
The source analysis also compares Fable 5.2 at High with Anthropic’s Claude Opus 5.5 at Max: Artificial Analysis lists scores of 62 for $9.08 per task and 59 for $10.04, respectively. This comparison suggests Fable may be competitive on that benchmark at High, but differences in tests, settings and actual workloads limit what buyers can infer. Teams should test representative tasks and track both quality and usage costs before changing deployment settings.
Products Worth Considering
LEVOIT Vital 200S-P Smart Air Purifier for Large Rooms up to 1875 ft², Quiet 27 dB Operation, AHAM VERIFIDE Certified, Air Quality Monitor,
Withings Sleep Under-Mattress Sleep Tracking Pad
BowFlex Results Series SelectTech Adjustable Dumbbells, Pair
As an affiliate, we earn on qualifying purchases.
How Fable’s Effort Settings Work
Anthropic released Claude Fable 5.2 on November 9, 2026, replacing Fable 5.1, according to the supplied source. It offers five effort levels, with Medium the default in Claude apps and Claude Code. Anthropic says the new model produces answers up to 30% faster than its predecessor. The source also reports an optional fast mode, priced at $16 per million input tokens and $80 per million output tokens; the benchmark cost table is not described as a fast-mode comparison.
The source describes Claude Mythos 5.2 as the same underlying model with fewer safety restrictions, available only to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Fable is open to users with safeguards for high-risk cybersecurity and research biology requests. These product and access descriptions are attributed to Anthropic in the source material, rather than independently established here.
Artificial Analysis’s Intelligence Index is the basis for the cost and score comparisons in this report. Its task costs are average API costs across the test, not a quote for an individual customer’s prompt. The source gives token rates and total output-token use across the index, which help explain why effort settings can produce different bills despite sharing the same listed per-token prices.
“The question for any business is not “is this the best model?” It is “how much thinking am I paying for, and do I need it?””
— Thorsten Meyer, ThorstenMeyerAI.com
What the Benchmark Cannot Establish
The supplied material does not provide enough detail to independently reproduce Artificial Analysis’s task-cost estimates or assess the full methodology behind each evaluation. The figures are described as averages across nine tests, so actual costs will depend on prompts, output length, usage patterns and settings. The source does not give a range of costs across repeated customer workloads.
It also remains unclear whether the benchmark score differences will translate into better results for particular business processes. Scores are rounded, and the source reports that High and Xhigh both scored 62. The FrontierCode result, where High slightly exceeded Max, is one test result; it does not establish that High is generally better. Anthropic’s separate claims about model capabilities and speed are company statements and are not validated by the benchmark figures alone.
The source says Anthropic recommends Opus 5.5 for most coding and knowledge work, but provides no full comparative evaluation across a broad range of customer tasks. Readers should treat the reported model rankings and costs as a snapshot of the named tests and date, not as a prediction of every user’s experience.
Testing Costs on Real Workloads
The next useful step for teams considering Fable 5.2 is to compare effort levels on their own representative tasks. That means recording output quality, completion time and API usage at more than one setting, then checking whether any improvement at a higher level justifies its added cost. Artificial Analysis’s published figures provide a starting point for that evaluation, not a substitute for it.
Further independent testing could clarify whether the score and cost pattern holds across different prompts, repeated runs and work outside the nine index evaluations. Anthropic may also publish additional model and pricing information. Until then, the confirmed comparison is limited to the released model’s stated token prices and the benchmark costs and scores reported in the source material.
Key Questions
How much does Fable 5.2 cost at each effort level?
Artificial Analysis reports estimated Intelligence Index costs per task of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. These are benchmark averages, not fixed prices for every request.
Does Max effort always produce the best result?
No. Max had the highest score in the supplied Intelligence Index table, but High and Xhigh both scored 62, and High slightly outscored Max on the cited FrontierCode 2.0 test. Results depend on the evaluation and task.
Why do costs change if the token rates are the same?
Anthropic’s listed input and output token rates are the same across the effort levels in the source. The benchmark cost estimates differ because the model used different amounts of output tokens while completing the index at each setting.
What is the difference between Fable 5.2 and Mythos 5.2?
The source describes Fable as the generally available model with safeguards for some high-risk cyber and biology requests. Mythos is described as a version with fewer such restrictions, available only to organizations accepted into Anthropic’s verification programs.
Should businesses switch to Fable 5.2?
The benchmark does not answer that for every organization. Teams can compare Fable and other models on their own tasks, measuring accuracy, response time and actual API costs before making a change.
Source: ThorstenMeyerAI.com
