🔍 Read the full analysis: The Price Of Effort: Claude Fable 5.2 Costs Per Task 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
Anthropic released Claude Fable 5.2 on November 9, 2026, and Artificial Analysis ranked its Max-effort setting first among 74 models the following day. The same evaluation puts average task costs between $3.10 at Low effort and $26.54 at Max, while the score rises from 54 to 65—making effort setting a major cost and performance choice.
Anthropic released Claude Fable 5.2 on November 9, and an Artificial Analysis evaluation published the next day ranked its Max-effort setting No. 1 among 74 models. The ranking comes with a wide price range: average cost per evaluation task rose from $3.10 at Low effort to $26.54 at Max, while the reported score increased from 54 to 65.
Artificial Analysis tested five effort settings on its Intelligence Index, which covers nine evaluations across agentic coding, business documents, science and knowledge. It reports an average API cost per task, not a universal price for every real-world use. At the default Medium setting, Fable 5.2 scored 56 and cost $3.80 per task. High scored 62 at $9.08; Xhigh also scored 62, at $14.83; Max scored 65 at $26.54.
The top score took the most time in the test: 46.4 minutes per task at Max, compared with 6.5 minutes at Low. Artificial Analysis’s figures also show total output tokens across the evaluation rising from 34.9 million at Low to 275 million at Max. Anthropic lists the same token prices for all effort levels—$8 per million input tokens and $40 per million output tokens—so the task-cost difference reflects, in part, how much the model generates while reasoning.
Anthropic says Fable 5.2 replaces Fable 5.1, writes answers up to 30% faster than its predecessor and offers an optional fast mode. The company lists fast-mode prices at $16 per million input tokens and $80 per million output tokens. These are company-provided product details; the benchmark costs above come from Artificial Analysis.
The Price of Effort Claude Fable 5.2 per task
Artificial Analysis ranked Max effort first among 74 models, but it also had the highest average task cost and longest run time. The setting is a trade-off, not a universal best choice.
Effort changes cost and score
Artificial Analysis measured five settings on its Intelligence Index; costs are evaluation averages, not fixed prices for every use.
| Effort | Score | Average task cost |
|---|---|---|
| Low | 54 | $3.10 |
| Medium | 56 | $3.80 |
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
The price of moving up
Higher effort can raise the benchmark score, but the extra cost does not always bring a higher displayed score.
$5.28 more per task for a six-point score increase.
$5.75 more per task; both displayed scores are 62.
Choose for the workload
A benchmark ranking cannot predict which setting will be best for a particular team or task.
- Max ranked first on the Intelligence Index.
- Low had the lowest reported average cost.
- High and Xhigh share the same displayed score.
- Compare cost, completion time and output quality on real tasks.
- Consider lower effort for routine work and test High for a stronger score.
- Do not assume Max improves every coding workload.
The Cost of Turning Effort Up
For organizations choosing a model setting, the results show a trade-off between benchmark score, cost and time, rather than one setting that leads on every measure. On this index, moving from Medium to High costs $5.28 more per task and raises the score by six points. Moving from High to Xhigh costs another $5.75, but the displayed score remains 62. Max adds three points over High and costs nearly three times as much.
Those differences matter for teams running many model tasks or waiting for results in interactive workflows. A benchmark’s average API cost is useful for comparing settings under the test’s conditions, but it is not a forecast of a company’s bill: actual use depends on prompts, output length and task mix. The longer Max run time may also matter where a person is waiting for an answer.
The comparison with Anthropic’s Claude Opus 5.5 offers another reference point in the same evaluation. Artificial Analysis lists Opus 5.5 at Max effort with a score of 59 and a $10.04 average task cost; Fable 5.2 at High scores 62 for $9.08. That result is specific to this index and does not establish that Fable will be less expensive or better for every workload.
Popular right now
KitchenAid Artisan Mini Tilt-Head Stand Mixer, 3.5 Qt, Contour Silver
Samsung 43-Inch M70H Series 4K Mini LED Smart TV (2026)
Samsung 48 4K OLED Smart TV (2026) with WireLogic 8ft Ultra High Speed HDMI Cable
As an affiliate, we earn on qualifying purchases.
Fable and Mythos Serve Different Users
Anthropic released two related models on November 9: Fable 5.2 for general users and Mythos 5.2 for organizations accepted into the company’s Cyber or Life Sciences Verification Programs. Anthropic describes Mythos as the same underlying model without some of Fable’s safety restrictions. Access to Mythos is limited to vetted organizations working in those dual-use areas.
Anthropic says Fable’s safeguards block some high-risk cybersecurity requests, including penetration testing and exploit generation, and most tasks in research biology. The company says blocked requests receive answers from an older model instead: Opus 4.8 for cybersecurity and Opus 5 for biology. Anthropic also reports that early testing found about 55% fewer blocks than with Fable 5.1; that figure is a company claim, not an independent measurement in the Artificial Analysis cost test.
Anthropic recommends that users start with Claude Opus 5.5 for most coding and knowledge work, then move to Fable when a task calls for it. The source material also describes striking prerelease demonstrations, including a Fermat’s Last Theorem proof effort involving roughly 1,000 agents. Those results are Anthropic’s account of its own testing, not independently verified evidence about typical customer performance.
Limits of the Cost Comparison
The reported costs are averages for tasks in Artificial Analysis’s evaluation, not prices guaranteed for an individual prompt or a company’s production workload. The supplied material does not detail how closely the test’s task mix or usage patterns match those of particular businesses. It also does not provide a direct measure of how often the extra Max-effort cost produces a meaningfully better outcome outside the benchmark.
Artificial Analysis displays the High and Xhigh scores as the same rounded value, 62; the figures provided do not show whether their unrounded scores differ. The source also says that High slightly outscored Max on one FrontierCode 2.0 coding test, but offers no broader evidence here for how effort settings compare across other coding workloads. Anthropic’s claims about safety performance, speed gains and prerelease achievements have not been independently verified in the supplied material.
Choosing Settings for Real Work
The immediate next step for teams is to compare effort levels on their own tasks, tracking cost, completion time and output quality rather than relying on a single index ranking. The evaluation suggests High may be worth testing where teams want a stronger score without Max’s reported expense, while lower settings may suit routine tasks. That is an interpretation of benchmark results, not a guarantee for any individual workflow.
Further independent testing could clarify how these cost and score patterns hold across different prompts and production conditions. The supplied material does not give a date for another Artificial Analysis evaluation or a subsequent Anthropic release. Anthropic’s access rules for Mythos and its recommendations for choosing models remain relevant to organizations considering sensitive cyber or biology work.
Key Questions
How much does Claude Fable 5.2 cost per task?
Artificial Analysis reports average costs on its Intelligence Index of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. These are evaluation averages, not fixed prices for every task.
Which effort setting ranked first?
Max effort scored 65 and ranked first among 74 models in Artificial Analysis’s November 10, 2026 results. It also had the highest reported cost and average task time among Fable 5.2’s five settings.
Is Max effort always better than High?
No. Max scored 65 versus High’s 62 on the Intelligence Index, but cost $26.54 per task compared with $9.08 at High. The source also says High scored slightly above Max on one FrontierCode 2.0 run, so the best setting may depend on the task.
What is the difference between Fable 5.2 and Mythos 5.2?
Anthropic says Fable is available to general users with safeguards for some high-risk cyber and biology requests. Mythos is described as the same underlying model without some of those restrictions and is limited to organizations approved through Anthropic’s verification programs.
Source: ThorstenMeyerAI.com
