TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Anthropic released Claude Fable 5.2 on November 9, and Artificial Analysis ranked it first among 74 models at Max effort the next day. The index score rose from 54 at Low to 65 at Max, while average task cost increased from $3.10 to $26.54.
Anthropic released Claude Fable 5.2 on November 9, and the model ranked first among 74 systems on the Artificial Analysis Intelligence Index at its Max effort setting a day later. That top score of 65 came with an average estimated cost of $26.54 per index task, compared with $9.08 and a score of 62 at High effort, putting a concrete price on the model’s adjustable reasoning settings.
Artificial Analysis evaluated five effort settings on its index, which covers nine evaluations in areas including agentic coding, business documents, science and general knowledge. Its reported task cost is the average API cost for one task in that test. Low scored 54 at $3.10 per task; Medium, the default in Claude apps and Claude Code, scored 56 at $3.80. High scored 62 at $9.08, Xhigh also scored 62 at $14.83, and Max reached 65 at $26.54. The ranking places High third, Xhigh second and Max first among the 74 models.
Completion time and token use also rose with effort. The average task took 6.5 minutes at Low and 46.4 minutes at Max. Across the full index, reported output-token use ranged from 34.9 million at Low to 275 million at Max. Anthropic lists Fable 5.2’s API prices at $8 per million input tokens and $40 per million output tokens at every effort setting; the different task bills reflect how much the model generates while working. An optional fast mode is listed at $16 per million input tokens and $80 per million output tokens.
The release includes Claude Mythos 5.2, which Anthropic describes as the same underlying model without Fable’s safeguards for certain high-risk cyber and biology work. Mythos access is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Anthropic says Fable’s safeguards can block high-risk requests and route them to older models; the company reported about 55% fewer blocks in early testing than with Fable 5.1.
Claude Fable 5.2 effort has a price
Higher effort raises benchmark scores, task costs and completion times. The best setting depends on the work and the value of faster, stronger results.
Five effort settings
Artificial Analysis measured average cost and score across its index tasks.
| Effort | Score | Cost per task |
|---|---|---|
| Low | 54 | $3.10 |
| Medium | 56 | $3.80 |
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
What the extra effort buys
Compared with High, Max adds three index points at almost three times the cost.
- Score: 62
- Cost: $9.08 per task
- Average time: not reported
- Score: 65
- Cost: $26.54 per task
- Average time: 46.4 minutes
Choose for the workload
Use your own tasks to weigh cost, elapsed time and output quality.
Match effort to the complexity and importance of the work.
Track cost per completed task, time and output quality.
Benchmark scores do not guarantee better results for a particular workflow.
Two models, different access
Anthropic describes Mythos 5.2 as the same underlying model without certain Fable safeguards.
Claude Fable 5.2
Safeguards can block high-risk requests and route them to older models.
Claude Mythos 5.2
Access is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs.
The Price of Extra Reasoning
The figures give businesses a way to weigh benchmark performance against cost and delay. In this index, moving from Medium to High adds six score points for $5.28 more per task. Moving from High to Xhigh raises the reported cost by about 63% without changing the rounded score. Max adds three points over High but costs almost three times as much and takes an average of 28.5 minutes longer per task.
Those differences can matter when a team runs many model tasks or needs results quickly. A higher index score does not, by itself, show that a more expensive setting will improve a particular company’s work. Results depend on task type, and the supplied source reports one coding evaluation where High slightly outscored Max: Fable 5.2 scored 49.1% versus 48.8% on Artificial Analysis’s FrontierCode 2.0 run.
The price comparison with Claude Opus 5.5 also complicates the “top model” label. On the index, Opus 5.5 at Max scored 59 for $10.04 per task, while Fable 5.2 at High scored 62 for $9.08. That gives buyers a specific comparison point, though it remains a benchmark comparison rather than a guarantee of real-world savings or better results.
Products Worth Considering
Apple MacBook Air 13.6-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Midnight
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
LG UltraGear 25G590B-B 25-Inch Full HD 1000Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
Two Models, Different Access
Fable 5.2 and Mythos 5.2 replace the corresponding 5.1 models, according to Anthropic. The company says the new models produce answers up to 30% faster than the 5.1 versions. That release claim is separate from the Artificial Analysis timing figures, which measure task completion on its index.
Anthropic recommends that users begin with Claude Opus 5.5 for most coding and knowledge work, moving to Fable when tasks call for it. Its own reported FrontierCode 2.0 results show Fable at High scoring 45.3%, close to Opus 5.5’s reported best of 45.5%; Anthropic described Fable’s cost per task in that comparison as “roughly half.” Artificial Analysis’ index comparison is a separate test and reports different scores and costs.
Anthropic also publicized results from prerelease work, including designing electronic devices from a short prompt and solving research problems. Those are company-reported examples, not independent findings in the supplied material. The release material does not establish how often ordinary users will see similar results.
““for roughly half the cost per task.””
— Anthropic, on its FrontierCode 2.0 comparison
Limits of the Ranking Data
The supplied figures do not explain whether Artificial Analysis’ index results predict performance on any particular business workflow. They also do not provide confidence intervals or enough detail to judge how stable the small score differences are, especially where scores are rounded. A three-point index gap should not be read as a measured improvement in every use case.
The source does not state whether the average task costs include all charges a customer might incur in practice, or how often users would need to retry or review outputs. Anthropic’s release claims about faster answers, fewer blocks and prerelease achievements are company statements; the material provided does not include independent verification of those claims. Access criteria and the number of organizations accepted for Mythos are also not specified.
Watch Real-World Costs
Businesses weighing the settings can compare them on their own workloads, tracking cost per completed task, elapsed time and output quality rather than relying on the index score alone. The figures available on November 10 establish the benchmark trade-off, but do not show how usage or pricing will change after the release.
Further details on Mythos access, independent checks of Anthropic’s release claims and results across customer tasks could clarify how the two-model approach performs beyond the published tests. Until then, the available ranking shows that Max buys the highest score on this index at a substantially higher cost and wait time than High.
Key Questions
What is the highest-ranked Fable 5.2 setting?
Max effort ranked first among 74 models on the Artificial Analysis Intelligence Index, with a score of 65 in figures accessed November 10, 2026.
How much does a Max-effort task cost?
Artificial Analysis reported an average API cost of $26.54 per index task at Max. The figure applies to its test, not necessarily to every user task.
Which setting offers a lower-cost alternative to Max?
High effort scored 62 for $9.08 per index task, compared with Max’s score of 65 for $26.54. The best setting depends on the work being done.
Who can use Claude Mythos 5.2?
Anthropic says Mythos is available to organizations accepted into its Cyber or Life Sciences Verification Programs. The supplied material does not specify how many organizations have access.
Source: Thorsten Meyer AI
