TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Anthropic released Claude Fable 5.2 on November 9, 2026, and Artificial Analysis reports that its five effort settings produce different scores, costs and completion times. In the index test, average cost per task ranged from $3.10 at Low to $26.54 at Max; the source says the ranking and measurements were accessed November 10.
Anthropic released Claude Fable 5.2 on November 9, 2026, and figures from Artificial Analysis show how its five effort settings trade off performance, cost and time. On the index cited in the source, scores range from 54 at Low to 65 at Max, while average cost per task ranges from $3.10 to $26.54.
Artificial Analysis’s Intelligence Index combines nine evaluations covering agentic coding, business documents, science and knowledge. Its reported Fable 5.2 results place Max first among 74 models, with a score of 65. Xhigh and High score 62, while Medium scores 56 and Low scores 54. These are the figures reported in the source, accessed November 10, 2026; they describe performance on that index, not a guarantee for every customer task.
The reported average cost per index task rises with effort: $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. Average time per task ranges from 6.5 minutes at Low to 46.4 minutes at Max. The source attributes these differences to how many output tokens the model uses for the full index: 34.9 million at Low and 275 million at Max.
Anthropic’s listed API rates are the same at every effort level: $8 per million input tokens and $40 per million output tokens. The source says an optional fast mode costs $16 per million input tokens and $80 per million output tokens, with token generation up to 2.5 times faster. Anthropic also says Fable 5.2 writes answers up to 30% faster than Fable 5.1. Those speed and pricing details are company statements; actual bills depend on usage.
Choose effort for the task and the budget
Artificial Analysis reports that higher effort can raise benchmark scores, costs and completion times. These index results are a reference point, not a forecast for every customer task.
Effort level tradeoffs
Reported averages show how cost and score change across the five settings.
| Effort | Index 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 comparison suggests
More effort does not guarantee a better result on every task.
- High scores 62 for $9.08 per index task.
- Moving from Medium to High adds six score points and $5.28 in reported average cost.
- Xhigh costs about 63% more than High, with the same rounded score.
- Max costs almost three times as much as High and takes more than twice as long on average.
Rates and total usage
Anthropic lists the same API token rates across effort levels; total bills depend on usage.
Fixed listed rates
$8 per million input tokens and $40 per million output tokens at every effort level.
Higher listed rates
$16 per million input tokens and $80 per million output tokens; token generation is stated to be up to 2.5 times faster.
Choose a setting for your work
The benchmark is a starting reference; your tasks and review needs determine the useful tradeoff.
Compare effort levels on work your team actually handles.
Measure cost, elapsed time and result quality together.
Use higher effort when deeper reasoning is worth its added cost and wait.
How Effort Changes the Bill
The results give teams a concrete reason to choose effort settings by task. In the index, moving from Medium to High adds six score points and $5.28 to the reported average task cost. That may make High attractive for work where a stronger result is worth the extra expense, though the index does not establish that the same gain will occur on a company’s own workload.
Beyond High, the reported tradeoff becomes less straightforward. Xhigh costs about 63% more than High while both have a rounded score of 62. Max adds three points over High, but costs almost three times as much and takes more than twice as long on average. The source also reports a FrontierCode 2.0 result in which Fable 5.2 at High scored 49.1%, slightly above its Max result of 48.8%. Together, these figures caution against assuming that the most expensive setting will improve every task.
For organizations budgeting API use, the distinction is practical: the per-token rate stays fixed, but the amount of model output can change substantially. Effort level affects total usage, and longer runs may matter as much as the nominal token price when people or systems are waiting for a result.
Products Worth Considering
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
Jackery Explorer 300 Portable Power Station with SolarSaga 100W Solar Panel
Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C – Graphite
As an affiliate, we earn on qualifying purchases.
Fable, Mythos and Access
The source describes a two-name release: Fable 5.2 is available to general users, while Mythos 5.2 is described as the same underlying model with fewer safety restrictions, limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. This access description comes from the source’s account of Anthropic’s release; the supplied material does not independently establish how the programs assess applicants.
Anthropic says Fable’s safeguards block high-risk cybersecurity requests, including penetration testing and exploit generation, as well as most tasks in research biology. The source says blocked cybersecurity prompts receive answers from Opus 4.8 and blocked biology prompts from Opus 5. Anthropic also reports that early testing found about 55% fewer blocks than with Fable 5.1. The source does not provide the test size or methodology for that comparison.
The article compares Fable with Anthropic’s Opus 5.5 as well. It reports that Opus 5.5 at Max scored 59 on the index for $10.04 per task, while Fable at High scored 62 for $9.08. Anthropic’s own FrontierCode 2.0 results, as quoted in the source, put Fable at High at 45.3%, close to Opus 5.5’s best reported 45.5%. These comparisons come from different tests and should not be treated as a universal ranking for business use.
What the Benchmarks Leave Open
The reported rankings and task costs are benchmark measurements, not observed spending or performance across customers’ real-world workloads. The source gives average cost per task for Artificial Analysis’s index but does not provide the per-task range, the underlying prompt mix, or enough detail to estimate how closely a particular business’s requests would match the test.
Some headline claims in the source come from Anthropic’s own prerelease accounts. The company reportedly described the model designing electronic devices, solving open problems in black-hole physics and formalizing a proof of Fermat’s Last Theorem with a large team of agents. The supplied material does not include independent verification or detailed methods for those examples. It also ends during its discussion of Anthropic’s coding results, so further context for that comparison is unavailable here.
It remains unclear how often users will see the full cost and time differences reported across effort levels, how Mythos access decisions will work in practice, and whether the reported safety-block rate holds across a broader set of requests. The source says neither Anthropic nor Artificial Analysis reviewed or sponsored its article, and discloses that it was drafted with Claude’s help.
Choosing a Setting for Work
Businesses can start by comparing the effort levels on representative tasks and tracking cost, elapsed time and result quality together. The index figures provide a reference point, but each team’s prompts, review requirements and tolerance for waiting will affect which setting makes sense. A task that benefits from deeper reasoning may justify High or Max; routine requests may not.
Anthropic’s stated guidance is to begin with Opus 5.5 for most coding and knowledge work, then consider Fable when a task needs it. Organizations interested in Mythos must meet Anthropic’s verification program criteria, though the supplied source does not give a timeline for applications or further access changes. Anthropic may publish more information about the model and its evaluations, but no specific next release or milestone is stated in the source.
Key Questions
What does each Fable 5.2 effort level cost per index task?
Artificial Analysis reports averages of $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. These figures apply to its Intelligence Index test, not every user request.
Which effort level scored highest?
Max scored 65 and ranked first among 74 models on the Artificial Analysis Intelligence Index cited in the source. High and Xhigh each scored 62.
Are token prices higher at higher effort levels?
The source reports the same listed API rates at every effort level: $8 per million input tokens and $40 per million output tokens. Total costs differ because the settings can use different amounts of output.
Who can use Mythos 5.2?
According to the source’s account of Anthropic’s release, Mythos is limited to organizations accepted into the company’s Cyber or Life Sciences Verification Programs. The supplied material does not detail application criteria.
Source: Thorsten Meyer AI
