🔍 Read the full analysis: A Task-by-Task Guide To Claude Fable 5.2 Effort-Level Costs on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Artificial Analysis measured Claude Fable 5.2 across five effort settings, with scores ranging from 54 at Low to 65 at Max and average costs per index task ranging from $3.10 to $26.54. The figures suggest High may offer a lower-cost alternative to Max for some workloads, but results come from a specific benchmark and do not establish what every business task will cost or achieve.
Artificial Analysis has published effort-level results for Anthropic’s Claude Fable 5.2, showing its Intelligence Index score rising from 54 at Low effort to 65 at Max while average cost per benchmark task increases from $3.10 to $26.54. The comparison gives organizations a task-specific view of the trade-off between benchmark performance, time and API spending, rather than a single price for using the model.
The independent model-ranking site tested Fable 5.2 at Low, Medium, High, Xhigh and Max on its Intelligence Index, which covers nine evaluations spanning agentic coding, business documents, science and knowledge. Artificial Analysis reports average task costs of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. The corresponding scores were 54, 56, 62, 62 and 65. Its reported times per task ranged from 6.5 minutes at Low to 46.4 minutes at Max.
The results show the largest score gain between Medium and High: six index points for $5.28 more per task, based on the reported averages. Xhigh cost more than High but had the same rounded score of 62. Max scored three points above High, with an average task cost close to three times as much. These are benchmark averages, not a quote for a particular customer’s workload.
Artificial Analysis also reports that Fable 5.2 at High scored 62 for $9.08 per task, while Claude Opus 5.5 at Max scored 59 for $10.04. Anthropic lists Fable’s standard API rates at $8 per million input tokens and $40 per million output tokens; those rates do not vary by effort setting. The source says higher effort can increase the output tokens used, which helps explain why task costs differ.
Claude Fable 5.2 Effort-Level Costs
Artificial Analysis compared five effort settings, with higher benchmark scores generally costing more per task. These averages are a comparison point, not a quote for your workload.
Five settings, rising costs
Scores and average costs reported for Artificial Analysis’s Intelligence Index tasks.
| Effort | Score | Avg. cost |
|---|---|---|
| Low | 54 | $3.10 |
| Medium | 56 | $3.80 |
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
What the comparison suggests
High may be a middle ground
High scored 62 for $9.08 per task, while Max scored 65 for $26.54. Test whether the extra score matters for your work.
Max is not always ahead
On FrontierCode 2.0, High scored 49.1% and Max scored 48.8%. Results can vary by evaluation.
API rates versus task costs
- $8 per million input tokens
- $40 per million output tokens
- Rates do not change by effort setting
- Vary across effort settings
- Higher effort can use more output tokens
- Do not predict every customer’s bill
Test before choosing
Measure the trade-off on tasks that reflect your actual workflows.
Compare effort settings on work your team actually performs.
Record quality, completion time, token use and cost.
Check whether High is sufficient where you might otherwise choose Max, and whether simpler work suits Low or Medium.
Choosing an Effort Level by Task
The comparison matters for teams weighing quality against latency and API expense. A higher-effort setting can improve results on some evaluations, but the benchmark does not show a uniform gain across every task. On Artificial Analysis’s FrontierCode 2.0 test, for example, the source reports Fable 5.2 scored 49.1% at High and 48.8% at Max. That result cautions against assuming the most expensive setting will always perform best.
For businesses, the practical question is whether a specific task benefits enough from additional model reasoning to justify its added cost and wait time. The published figures offer a comparison point, but they do not predict an organization’s bill: actual use depends on prompts, workload, token consumption and the chosen service settings.
Popular right now
Vitamix VX1 Professional-Grade Blender, 64 oz, Black
De'Longhi Magnifica Start Fully Automatic Espresso Machine with Milk Frother, Silver
Roborock RockNeo Q110H Wire-Free Robot Lawn Mower
As an affiliate, we earn on qualifying purchases.
Anthropic’s New Model Options
Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, replacing the prior 5.1 versions. The supplied source describes Fable as generally available and Mythos as restricted to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Anthropic says Fable has safeguards that block certain high-risk cybersecurity and research-biology requests; Mythos is intended for vetted organizations in those dual-use areas.
Anthropic says the models cost $8 per million input tokens and $40 per million output tokens, and that answers are written up to 30% faster than with the 5.1 versions. It also offers an optional fast mode at $16 per million input tokens and $80 per million output tokens. These are company-provided product details, distinct from Artificial Analysis’s benchmark task-cost estimates. The source also says Anthropic recommends starting with Claude Opus 5.5 for most coding and knowledge work, using Fable when a task calls for it.
Benchmark Limits and Open Questions
The supplied material does not include the full task-level results, test prompts or enough methodological detail to determine how closely the Intelligence Index represents any particular organization’s work. Scores are shown as rounded values, so identical reported scores at High and Xhigh do not prove the underlying results were exactly equal. The average cost and time figures also may not match a real deployment.
Anthropic’s reported performance claims and descriptions of safeguards are company statements. The source says its results on several demanding demonstrations came from prerelease testing, but those accounts are not independently verified in the material provided. It is also unclear how much the reported faster answer generation changes end-to-end task completion time for different workflows.
What Teams Can Measure Next
Organizations considering Fable can compare settings on their own representative tasks, tracking quality, completion time, token use and cost rather than relying on benchmark rank alone. A useful next test is whether High delivers acceptable results where Max would otherwise be selected, and whether simpler tasks can be handled at Low or Medium.
Further evidence would include independent replication, more detailed evaluation methods and real-world cost comparisons across different workloads. Until then, the Artificial Analysis figures provide a published benchmark comparison, not a guarantee of savings or performance in production.
Key Questions
How much does a Claude Fable 5.2 benchmark task cost?
Artificial Analysis reports average costs from $3.10 at Low effort to $26.54 at Max on its Intelligence Index. These are test averages, not fixed prices for every user task.
Which effort level had the highest reported score?
Max effort scored 65 on the reported Intelligence Index, ranking first among the five Fable 5.2 settings listed in the source. It also had the highest reported average cost and task time.
Is Max always better than High?
No. Max scored three points above High on the Intelligence Index, but the source reports High slightly ahead of Max on the FrontierCode 2.0 coding test. Results vary by evaluation.
Do Fable’s API rates change with effort level?
The source says the standard rates remain $8 per million input tokens and $40 per million output tokens. The benchmark’s task costs vary with token use, which the source says rises at higher effort levels.
Source: ThorstenMeyerAI.com
