🔍 Read the full analysis: How Much More Does Each Claude Fable 5.2 Effort Level Cost? on ThorstenMeyerAI.com
Get smart everyday buys 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’s Intelligence Index reports that Claude Fable 5.2 costs $3.10 per test task at Low effort and $26.54 at Max, while its score rises from 54 to 65. The figures come from one benchmark suite, not a universal price quote; actual API bills depend on token use and task.
Claude Fable 5.2 cost an average of $3.10 per task at Low effort and $26.54 at Max in Artificial Analysis’s Intelligence Index, while its displayed score rose from 54 to 65. The November 10 benchmark figures give businesses a concrete comparison of the model’s effort settings, but they are costs for tasks in that index—not a fixed charge for every real-world request.
Artificial Analysis tested five settings—Low, Medium, High, Xhigh and Max—on an index of nine evaluations covering areas including agentic coding, business documents, science and general knowledge. The reported score and average task cost moved together unevenly: Medium scored 56 for $3.80, High scored 62 for $9.08, Xhigh also scored 62 for $14.83, and Max scored 65 for $26.54. Scores are rounded as displayed by Artificial Analysis.
The measured time per task also rose with effort: 6.5 minutes at Low, 8.3 at Medium, 17.9 at High, 26.4 at Xhigh and 46.4 at Max. Artificial Analysis reported that the model generated 34.9 million output tokens across the index at Low, compared with 275 million at Max. Those aggregate token totals help explain why task costs differ even though the listed token prices are the same across effort settings.
According to the source material, Anthropic lists Fable 5.2 API pricing at $8 per million input tokens and $40 per million output tokens. The benchmark’s task costs reflect the tokens used in its evaluations, not a separate per-task price set by Anthropic. Anthropic also offers an optional fast mode at higher listed rates—$16 per million input tokens and $80 per million output tokens—according to its published pricing information as described in the source.
How Much More Does Each Claude Fable 5.2 Effort Level Cost? Cost, score and time by setting
Artificial Analysis measured higher average task costs as effort increased, while score gains varied. These are benchmark averages, not fixed prices for real-world requests.
Benchmark cost by effort
Average API cost per task in Artificial Analysis’s Intelligence Index.
| Effort | Cost per task | Score |
|---|---|---|
| Low | $3.10 | 54 |
| Medium | $3.80 | 56 |
| High | $9.08 | 62 |
| Xhigh | $14.83 | 62 |
| Max | $26.54 | 65 |
Time rises with effort
Measured average minutes per benchmark task.
Why costs differ
The listed token rates are the same across effort settings; benchmark usage changes with the tokens generated.
- 34.9 million output tokens across the index
- $3.10 average cost per task
- 275 million output tokens across the index
- $26.54 average cost per task
What teams should take from it
Higher effort can raise cost and latency without improving every measured result.
Test the prompts and workflows your team actually uses.
Record quality, latency and token-based cost across levels.
Use a higher setting when its gains justify added time and cost.
The Cost of Higher Effort
The results show why choosing an effort setting is a cost-and-latency decision, not simply a choice to buy a better answer. Moving from Medium to High added six displayed index points and $5.28 per task in this test. Moving from High to Xhigh raised measured cost by about 63% while leaving the rounded score at 62. Max added three points over High, but its benchmark task cost was nearly three times as high.
That difference could matter for teams running many requests or waiting on results interactively. A 46.4-minute average at Max is materially different from 6.5 minutes at Low, although benchmark timing does not guarantee the same duration for an individual task. The source also reports that Fable 5.2 scored 49.1% at High and 48.8% at Max on Artificial Analysis’s FrontierCode 2.0 coding test. That single evaluation is a reminder that higher effort did not improve every measured result.
In a comparison cited by the source, Fable 5.2 at High scored 62 for $9.08 per task, while Claude Opus 5.5 at Max scored 59 for $10.04 on the index. That makes Fable’s High setting a potentially useful benchmark option, but it does not establish that it will be cheaper or more capable for every company’s workload. Historical test results are not a guarantee of production performance or cost.
Popular right now
Dyson Airwrap i.d. Multi-Styler for Straight and Wavy Hair
Breville Bambino Espresso Machine, Brushed Stainless Steel
Vitamix VX1 Professional-Grade Blender, 64 oz, Black
As an affiliate, we earn on qualifying purchases.
How the Effort Test Works
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, according to the source. The models replace their 5.1 predecessors. Artificial Analysis’s figures, accessed the following day, rank Fable 5.2 at Max first among 74 models on its Intelligence Index, with Fable at Xhigh and High also appearing near the top. The ranking is one benchmark view, not a general measure of every use case.
The benchmark’s “cost per task” is the average API cost for one task in Artificial Analysis’s index. It should not be confused with a subscription charge, an estimate for a particular customer, or the cost of a complete business workflow. Actual usage depends on the task, input and output token counts, and selected settings. The source says Medium is the default effort level in Claude apps and Claude Code.
The source also describes a distinction between the two model names based on access and safeguards: Fable is available to users with protections for high-risk cyber and biology requests, while Mythos is described as available to organizations accepted into Anthropic’s verification programs. Those release details are separate from Artificial Analysis’s effort-level cost test, which concerns Fable 5.2.
Limits of the Cost Estimates
The published figures do not show what a particular business will pay for its own prompts, coding jobs or document workflows. Artificial Analysis’s per-task averages apply to its test suite, and the source does not provide a breakdown of costs for each individual evaluation. Nor does a rounded score reveal whether small differences between effort settings are meaningful for a specific application.
The source does not specify how representative the nine evaluations are of typical customer workloads, or whether the measured timings will match a given deployment. It also does not provide independent verification of Anthropic’s separate claims about Fable’s prerelease achievements. Those claims are not needed to establish the benchmark costs and should not be read as evidence that a normal user will see comparable results.
Choosing a Setting for Workloads
For teams considering Fable 5.2, the next practical step is to test representative tasks at more than one effort level and record quality, latency and token-based cost. A workflow that gains little from Max may be less expensive at High or Medium; tasks where a small improvement has high value may warrant a higher setting. These benchmark results can guide testing, but they cannot replace it.
Anthropic’s stated recommendation, as reported in the source, is to begin with Opus 5.5 for most coding and knowledge work and use Fable when the task calls for it. Future comparisons may clarify how the models perform on additional evaluations and customer workloads. Until then, the published figures support a measured conclusion: higher effort can raise both score and cost sharply, but the gains are not uniform.
Key Questions
How much does Claude Fable 5.2 cost at each effort level?
Artificial Analysis measured average costs of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max per task in its Intelligence Index. These are benchmark averages, not fixed charges for every request.
What score did Fable 5.2 get at Max effort?
Artificial Analysis displayed a score of 65 at Max effort, ranking it first among 74 models on its Intelligence Index as reported on November 10, 2026. The ranking applies to that index, not all possible tasks.
Does higher effort always improve performance?
No. On the Intelligence Index, Xhigh and High both displayed a score of 62, while Max scored 65. The source also reports a FrontierCode 2.0 result of 49.1% at High versus 48.8% at Max, showing that the highest setting did not lead on every test.
Why do effort settings cost different amounts if token prices are the same?
The listed per-token rates are unchanged across settings, but the amount of output generated can differ. Artificial Analysis reported 34.9 million output tokens across the index at Low and 275 million at Max, contributing to different average task costs.
Source: ThorstenMeyerAI.com
