TL;DR
Get everyday essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Artificial Analysis ranked Anthropic’s Claude Fable 5.2 first among 74 models on its Intelligence Index at Max effort, with a score of 65. The same model scored 62 at High for $9.08 per test task, compared with $26.54 at Max; the results come from an independent benchmark, not a measure of every team’s real-world performance.
Anthropic’s Claude Fable 5.2 ranked first among 74 models on the Artificial Analysis Intelligence Index at Max effort, scoring 65 in results accessed November 10. The benchmark’s average API cost was $26.54 per task at that setting, while High effort scored 62 at $9.08 per task, giving businesses a materially cheaper option with a slightly lower index score.
Artificial Analysis evaluates models across nine assessments covering agentic coding, business documents, science and knowledge. Its reported cost per task is an average API cost for one task in that index, not a guaranteed cost for an individual company’s workload. Scores are rounded as displayed by the benchmark.
Fable’s five effort settings produced a range of results: Low scored 54 at $3.10 per task and took an average of 6.5 minutes; Medium, the default in Claude apps and Claude Code, scored 56 at $3.80 and took 8.3 minutes. High scored 62 at $9.08, Xhigh also scored 62 at $14.83, and Max scored 65 at $26.54. Average task time at Max was 46.4 minutes.
Anthropic lists the same standard token prices across Fable’s effort levels: $8 per million input tokens and $40 per million output tokens. The source attributes the differing task costs to how many tokens the model uses while reasoning. Across the full index run, reported output-token use ranged from 34.9 million at Low to 275 million at Max. Anthropic also offers an optional fast mode at $16 per million input tokens and $80 per million output tokens, which it says can generate tokens up to 2.5 times faster.
Claude Fable 5.2 ranks first What each effort level costs
Artificial Analysis ranked Fable 5.2 first among 74 models at Max effort. Higher effort brought a higher index score, but also a higher average cost and longer task time.
Effort changes the trade-off
Benchmark averages show how score and API cost varied across Fable’s five effort settings.
| Effort | Score | Cost per task | Average time |
|---|---|---|---|
| Low | 54 | $3.10 | 6.5 minutes |
| Medium | 56 | $3.80 | 8.3 minutes |
| High | 62 | $9.08 | Not reported |
| Xhigh | 62 | $14.83 | Not reported |
| Max | 65 | $26.54 | 46.4 minutes |
More effort did not always win
The overall index and a coding test showed different patterns.
- High scored 62 for $9.08 per task.
- Max scored 65 for $26.54 per task.
- Xhigh and High had the same rounded score.
- High scored 49.1% on FrontierCode 2.0.
- Max scored 48.8% on that test.
- More effort did not improve this measured result.
Token prices stay the same
Anthropic lists standard token rates across effort settings; task costs vary with token use while reasoning.
$8 per million tokens
Standard API input price across effort levels.
$40 per million tokens
Standard API output price across effort levels.
Higher token rates
$16 per million input and $80 per million output tokens; Anthropic says it can generate tokens up to 2.5 times faster.
Put the benchmark in context
The index offers a comparison point, while your own workload determines the practical value.
- Its average task cost is not a guaranteed cost for a company’s workload.
- The index covers nine assessments, including coding, business documents, science and knowledge.
- Scores are rounded, and the supplied material does not show how closely the task mix matches a particular business.
- Anthropic recommends starting with Opus 5.5 for most coding and knowledge work, then moving to Fable when needed.
The Cost of the Top Score
The results make effort setting a practical budget choice for teams weighing model quality against cost and response time. In this benchmark, moving from Medium to High added six score points for $5.28 more per task. Xhigh cost about 63% more than High but had the same rounded score; Max added three points over High while costing nearly three times as much.
Those figures do not establish that Max is the right choice for demanding work in general. In Artificial Analysis’s FrontierCode 2.0 coding test, Fable 5.2 at High scored 49.1%, slightly above its 48.8% at Max. The benchmark results suggest teams may want to select effort by task and verify performance on their own work, since more effort did not improve every measured result.
The comparison with Anthropic’s Claude Opus 5.5 also gives buyers a reference point. Artificial Analysis lists Opus 5.5 at Max with a score of 59 and a cost of $10.04 per task. Fable at High scored 62 for $9.08 in the same index. These are benchmark comparisons; they do not guarantee a similar cost or advantage in a company’s own use.
Popular right now
De'Longhi Magnifica Duo Automatic Espresso Machine with Milk Frother – Silver
KitchenAid Artisan Mini 3.5-Qt Tilt-Head Stand Mixer, Candy Apple Red
Breville Barista Express Espresso Machine with Built-In Grinder, Black Sesame
As an affiliate, we earn on qualifying purchases.
Fable and Mythos Split
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, replacing the 5.1 versions. According to Anthropic, they share the same underlying model, but Fable includes safeguards for high-risk cybersecurity work and most research biology tasks. Mythos lacks those blocks and is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs.
Anthropic says blocked Fable queries receive responses from an older model: Opus 4.8 for cybersecurity and Opus 5 for biology. The company also reported about 55% fewer blocks in early testing than with Fable 5.1. These are company statements; the supplied material does not include independent verification of the testing or its methods.
Anthropic says the new models produce answers up to 30% faster than their 5.1 counterparts. It recommends starting with Claude Opus 5.5 for most coding and knowledge work and moving to Fable when a task needs it. The company’s accounts of prerelease accomplishments, including device design and mathematical work, are also Anthropic claims and do not establish typical customer results.
“For most coding and knowledge work, start with Claude Opus 5.5 and move up to Fable only when the task needs it.”
— Anthropic
Limits of the Benchmark Results
The index provides a standardized comparison, but the supplied source does not give enough detail to determine how closely its task mix or average costs match any particular business’s usage. Real-world costs depend on the work performed, including token use and the chosen effort setting. The displayed scores are rounded, so equal scores at High and Xhigh do not show whether their underlying results were identical.
The source also does not establish how Fable and Mythos perform across customer deployments, or independently verify Anthropic’s claims about safety blocks, speed and prerelease achievements. The benchmark ranking measures performance on its selected assessments; it does not settle which model is best for every use case.
Test Settings on Real Workloads
Teams considering Fable can compare Low, Medium and High on representative tasks, recording both quality and API spend before deciding whether higher settings are worth the additional cost and wait time. Artificial Analysis’s figures offer a starting point for that evaluation, while Anthropic’s recommendation points most coding and knowledge work users toward Opus 5.5 first.
Further comparisons could clarify how the models perform on specific business workloads and whether the benchmark pattern holds outside the index. No next benchmark release or additional evaluation date is specified in the supplied source.
Key Questions
What ranked first on the Artificial Analysis Intelligence Index?
Claude Fable 5.2 at Max effort ranked first among 74 models, scoring 65 in results accessed November 10, 2026.
How much does each Fable 5.2 effort level cost per index task?
Artificial Analysis reported average costs 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 task.
Is Max effort always better than High?
No. Max scored higher on the overall index, but High and Max tied at 62 on the rounded index score at High and Xhigh? Specifically, High and Xhigh both scored 62, while Max scored 65. On FrontierCode 2.0, the source reports High at 49.1% and Max at 48.8%.
What is the difference between Fable 5.2 and Mythos 5.2?
Anthropic describes them as the same underlying model with different access and safety controls. Fable is available with safeguards for certain high-risk cyber and biology tasks; Mythos is for organizations accepted into Anthropic’s verification programs.
Source: Thorsten Meyer AI
