🔍 Read the full analysis: Comparing Claude Fable 5.2’S Effort Levels And What Each Task Costs on ThorstenMeyerAI.com
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
TL;DR
Artificial Analysis’ November 10 comparison puts Claude Fable 5.2’s five effort levels between $3.10 and $26.54 per benchmark task, with scores ranging from 54 to 65. High effort scored 62 for $9.08, while Max scored 65 for $26.54; the results describe one benchmark, not a guaranteed cost or outcome for everyday tasks.
Artificial Analysis’ November 10 figures show that Claude Fable 5.2’s benchmark cost varies widely by effort setting: its Intelligence Index score rises from 54 at Low for $3.10 per task to 65 at Max for $26.54. The comparison gives businesses a concrete measure of the cost and time trade-off in Anthropic’s model, but it reflects a particular test suite rather than a universal price per real-world request.
Artificial Analysis tested Fable 5.2 at five settings—Low, Medium, High, Xhigh and Max—on its Intelligence Index, which covers nine evaluations spanning agentic coding, business documents, science and knowledge. It reports scores of 54, 56, 62, 62 and 65, respectively. The corresponding average API costs per task were $3.10, $3.80, $9.08, $14.83 and $26.54. Medium is the default in Claude apps and Claude Code, according to the source report.
The reported average 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’ whole-index output-token totals ranged from 34.9 million at Low to 275 million at Max. Anthropic lists the same standard rates across effort levels—$8 per million input tokens and $40 per million output tokens—so the benchmark’s differing task costs are tied to differing token use, including the model’s reasoning output.
High scored 62 for $9.08, while Max scored 65 for $26.54. Xhigh cost $14.83 and also scored 62 when rounded as shown by Artificial Analysis. In a separate coding evaluation cited in the source report, Fable 5.2 scored 49.1% at High and 48.8% at Max on FrontierCode 2.0. These benchmark results do not establish that High will outperform Max on other tasks.
Claude Fable 5.2 Effort and cost
Higher effort scores more on this benchmark, but also costs more and takes longer. The results can inform budgets, not predict the cost or outcome of every real-world task.
Five effort settings
Artificial Analysis tested nine evaluations across coding, business documents, science and knowledge.
| Effort | Score | Avg. cost | Avg. time |
|---|---|---|---|
| Low | 54 | $3.10 | 6.5 min |
| Medium | 56 | $3.80 | 8.3 min |
| High | 62 | $9.08 | 17.9 min |
| Xhigh | 62 | $14.83 | 26.4 min |
| Max | 65 | $26.54 | 46.4 min |
The cost of more effort
The benchmark shows a growing spend and runtime as effort rises, while displayed scores do not increase at every step.
What teams should weigh
Choose effort against the needs and costs of the actual workflow.
- Higher benchmark scores at greater effort
- High scored 62 for $9.08 per task
- Teams can compare quality, token use, latency and API cost on representative work
- Benchmark averages are not fixed prices for requests
- Actual cost depends on task, token use and configuration
- The index does not measure review time or the cost of errors
The Cost of More Model Effort
The figures make the effort setting a practical budget decision, not just a model preference. On this test, moving from Medium to High added six index points and raised the average task cost from $3.80 to $9.08. Moving from High to Max added three points while increasing the reported cost by $17.46 per task and the average run time by 28.5 minutes.
That difference may matter for organizations running many tasks or waiting for a person to review each result. A higher benchmark score may justify extra expense for some difficult work, but the index does not show that every business prompt benefits from Max. Actual costs will depend on the task, how many tokens the model uses and the application’s configuration. The reported figures are benchmark averages, not a fixed quote for every request.
The comparison also cautions against treating the top index rank as a simple answer to which setting to use. Artificial Analysis placed Fable 5.2 at Max first among 74 models, but its High setting ranked third and Xhigh second in the same reported comparison. Those rankings are specific to the index and its scoring and cost methodology.
Products Worth Considering
LG UltraGear 34GX90SA-W 34-inch WQHD Curved OLED Gaming Monitor
LG UltraGear 27G640A-B 27-Inch QHD IPS Gaming Monitor, 300Hz, 1ms, G-SYNC Compatible, FreeSync Premium, DisplayHDR 400
LG UltraGear 32GX870A-B 32-inch 4K OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
Fable 5.2 and Anthropic’s Release
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, replacing the respective 5.1 versions, according to the source report. Anthropic says the models generate answers up to 30% faster than their predecessors. It also offers an optional fast mode, which the company says generates tokens up to 2.5 times faster, at listed rates of $16 per million input tokens and $80 per million output tokens.
The report describes the distinction between Fable and Mythos as one of access and safeguards. Fable is available to users with protections against certain high-risk cybersecurity and research-biology requests. Anthropic says Mythos is for organizations accepted into its Cyber or Life Sciences Verification Programs. The source material says blocked Fable requests may be answered by an older model, but does not provide independent testing of these safeguards.
Anthropic recommends starting with Claude Opus 5.5 for most coding and knowledge work, then using Fable when a task calls for it, according to the source report. The comparison cited there puts Opus 5.5 at Max at a score of 59 and $10.04 per index task, against Fable 5.2 at High with a score of 62 and a $9.08 task cost. These are Artificial Analysis results, not a guarantee that Fable will be cheaper or score higher on a particular customer’s work.
“The new models write their answers up to 30% faster than the 5.1 versions.”
— Anthropic, as reported by ThorstenMeyerAI.com
Limits of the Benchmark Comparison
The supplied figures do not establish how Fable 5.2 will score or cost on any one company’s workload. Artificial Analysis’ task averages come from its nine-evaluation Intelligence Index; the source material does not give task-by-task cost distributions or explain how closely those tasks match typical customer use. Token consumption and cost can differ between prompts.
The index scores are shown as rounded values, so the identical displayed score of 62 at High and Xhigh does not prove that the underlying results were exactly equal. The source report also cites separate Anthropic claims about prerelease achievements, including a Fermat’s Last Theorem project, but says those figures were not independently verified. They are not evidence of typical performance or part of the per-task cost comparison.
It is also unclear from the supplied information how the benchmark’s average costs would change under optional fast mode, or how the model’s safeguards perform across a broad range of real-world requests. The source material identifies Anthropic and Artificial Analysis as the sources of the release and index data; it does not provide an independent audit of the underlying benchmark runs.
What Teams Can Measure Next
Organizations considering Fable can compare settings on their own representative tasks, tracking quality, token use, latency and API cost rather than relying only on the index rank. That would show whether High’s reported benchmark balance or a more expensive setting fits a specific workflow. Teams should also account for review time and the cost of errors, which the index figures alone do not measure.
The source material gives no scheduled follow-up benchmark or further release date. For now, the next relevant information would be independent or customer evaluations across real workloads, plus clearer comparisons of fast-mode performance and cost. Until then, the published cost figures should be read as Artificial Analysis averages for its test suite, not as a forecast for an individual deployment.
Key Questions
How much does a Fable 5.2 task cost at each effort level?
Artificial Analysis reports average costs per Intelligence Index task of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. The amounts are benchmark averages, not fixed prices for every user request.
Which effort level scored highest?
Max scored 65 on Artificial Analysis’ Intelligence Index, the highest result in the five-setting comparison. The index score does not show that Max is best for every task.
Is High a better value than Max?
In this index, High scored 62 for $9.08 per task, while Max scored 65 for $26.54. Whether the extra score is worth the additional cost depends on the work; the benchmark cannot settle that for every use case.
Why do costs vary if the listed token prices are the same?
Anthropic’s reported standard rates are $8 per million input tokens and $40 per million output tokens across the effort levels. Artificial Analysis reports different task costs because the settings use different amounts of tokens, including reasoning output.
Are the reported costs guaranteed for real-world tasks?
No. They are average API costs on Artificial Analysis’ Intelligence Index. A real request’s cost can vary with its prompt, task and token use.
Source: ThorstenMeyerAI.com
