🔍 Read the full analysis: Claude Fable 5.2: Which Effort Level Gives You The Right Cost Per Task? on ThorstenMeyerAI.com
TL;DR
Anthropic released Claude Fable 5.2 on Nov. 9, 2026, with five selectable effort levels. Artificial Analysis measured scores from 54 to 65 and average test-task costs from $3.10 at Low to $26.54 at Max; its results show that higher effort does not always improve performance enough to justify the added cost.
Anthropic released Claude Fable 5.2 on Nov. 9, offering five effort settings that trade off test performance, cost and time. In figures published by Artificial Analysis, average cost per Intelligence Index task ranged from $3.10 at Low to $26.54 at Max, while scores rose from 54 to 65—results that give organizations a concrete way to weigh additional model reasoning against its expense.
Artificial Analysis tested Fable 5.2 on its Intelligence Index, a set of nine evaluations covering agentic coding, business documents, science and knowledge. The firm reported a score of 54 at Low, 56 at Medium, 62 at High, 62 at Xhigh and 65 at Max. The corresponding average cost per test task was $3.10, $3.80, $9.08, $14.83 and $26.54. Artificial Analysis lists Medium as the default in Claude apps and Claude Code.
The results show the sharpest score increase between Medium and High: six points for $5.28 more per task. Xhigh cost about 63% more than High but received the same rounded score of 62. Max scored three points above High, while costing nearly three times as much. Average task time also rose with effort, from 6.5 minutes at Low to 46.4 minutes at Max.
Anthropic says Fable 5.2 replaces Fable 5.1 and offers answers up to 30% faster than its predecessor. The company lists API prices of $8 per million input tokens and $40 per million output tokens at every effort setting; the difference in the test’s task costs reflects how many tokens the model uses, including tokens spent reasoning. An optional fast mode, Anthropic says, generates tokens up to 2.5 times faster and costs $16 per million input tokens and $80 per million output tokens.
Claude Fable 5.2 Choosing effort by task
Five effort levels trade off benchmark score, task cost and time. The right setting depends on whether extra performance is worth the added expense for your work.
Effort changes the trade-off
Artificial Analysis measured nine evaluations; these benchmark averages may differ from customer workloads.
| 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 |
Where added effort pays off
High scored six points above Medium for $5.28 more per task in the benchmark.
- Score: 62
- Average cost: $9.08
- Average task time: 17.9 minutes
- Score: 65
- Average cost: $26.54
- Average task time: 46.4 minutes
Match the setting to the work
Medium is the default in Claude apps and Claude Code; test settings on representative tasks before scaling.
Use lower effort for work that may not need the highest benchmark score.
Measure task quality, API usage and waiting time on your own prompts.
Check whether Max’s added score justifies its higher cost and longer task time.
Access and safety differ
Safeguards included
Anthropic says Fable applies safeguards to high-risk cybersecurity requests and most research-biology tasks.
Restricted access
Anthropic describes Mythos as the same underlying model without some restrictions, available to approved Cyber or Life Sciences Verification Program organizations.
How Effort Changes the Bill
The figures matter for teams deciding how to deploy a costly model across routine and demanding work. One setting does not fit every task: the Intelligence Index results suggest High may offer a different cost-performance balance from Max, while Low and Medium cost less for workloads that may not need the highest score. These are averages from a particular benchmark, not a price quote for every prompt or business workflow.
Time is another part of the trade-off. A test task taking 46.4 minutes at Max rather than 17.9 minutes at High could affect how quickly a person or automated system receives results. That does not establish how long real-world tasks will take, but it indicates that effort settings can change both API usage and waiting time.
The benchmark also cautions against assuming that more reasoning always means better results. In Artificial Analysis’s FrontierCode 2.0 agentic coding test, High scored 49.1%, slightly above Max at 48.8%, according to the source material. The result is specific to that test; it does not show that High will outperform Max on other tasks.
Popular right now
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
Bose QuietComfort Earbuds (2nd Gen) – True Wireless Bluetooth Earbuds with Active Noise Cancellation
Vitamix 5200 Blender with 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
Fable, Mythos and Safety Controls
Anthropic introduced Fable 5.2 and Mythos 5.2 as related models with different access and safety settings. According to Anthropic, Fable is available to users with safeguards for high-risk cybersecurity requests and most research-biology tasks. Queries blocked by those safeguards may be answered by an older model instead: Opus 4.8 for cybersecurity and Opus 5 for biology.
Anthropic says Mythos 5.2 is the same underlying model without some of those safety restrictions, and is available only to organizations accepted into its Cyber or Life Sciences Verification Programs. The company also says early testing found about 55% fewer blocks with Fable 5.2 than with Fable 5.1. That is a company-reported result, not an independently verified finding in the supplied material.
Anthropic recommends that users start with Claude Opus 5.5 for most coding and knowledge work, moving to Fable when a task requires it. In Artificial Analysis’s comparison, Opus 5.5 at Max scored 59 at $10.04 per task, while Fable 5.2 at High scored 62 at $9.08. Those results are tied to the benchmark and its reported settings; they do not establish which model is better for a particular organization.
„For most coding and knowledge work, start with Claude Opus 5.5 and move up to Fable only when the task needs it.“
— Anthropic
Benchmark Limits and Open Questions
The Artificial Analysis figures describe performance on its nine-evaluation Intelligence Index, not the full range of work customers may run. The supplied results do not establish whether the reported cost per task will match a company’s own prompts, tool use, output lengths or volume. Scores are shown as rounded by Artificial Analysis, so equal rounded scores at High and Xhigh do not necessarily mean identical underlying results.
Anthropic also described prerelease accomplishments, including work on electronic devices and a large agent-assisted mathematical proof effort. Those accounts are company claims, not independently verified results in the source material. They should not be treated as typical performance for ordinary users. The available information does not establish how often customers will need Max, or how its added cost and time will affect production workloads.
Choosing Settings in Real Workflows
Organizations considering Fable can compare effort levels on their own representative tasks, tracking cost, latency and task quality rather than relying on an index score alone. The published benchmark suggests that High is worth testing against Max for work where the extra three index points may not justify nearly triple the reported task cost. For less demanding workloads, Anthropic’s recommendation is to begin with Opus 5.5.
Further comparisons may clarify whether the benchmark’s cost and performance pattern holds across different applications and over longer use. For now, the confirmed public figures are Artificial Analysis’s test results and Anthropic’s stated pricing, model-access rules and performance claims; customer-specific outcomes remain to be measured.
Key Questions
What are Claude Fable 5.2’s effort settings?
Fable 5.2 offers Low, Medium, High, Xhigh and Max. Anthropic lists Medium as the default in Claude apps and Claude Code.
Which effort level had the lowest reported cost?
Artificial Analysis reported an average cost of $3.10 per Intelligence Index task at Low. The figure comes from that benchmark and may not match the cost of a specific customer workload.
Is Max worth the added cost over High?
In Artificial Analysis’s results, Max scored 65 versus 62 at High, with average task costs of $26.54 and $9.08, respectively. Whether the additional score is worth the cost depends on the task; the benchmark does not answer that for every use case.
How is Mythos 5.2 different from Fable 5.2?
Anthropic describes Mythos as the same underlying model without some of Fable’s safety restrictions. Access is limited to organizations accepted into the company’s Cyber or Life Sciences Verification Programs; Fable includes safeguards for certain high-risk requests.
Do higher effort settings always produce better results?
No. The Intelligence Index score rose overall from Low to Max, but Artificial Analysis reported High slightly ahead of Max on its FrontierCode 2.0 coding test. Results depend on the evaluation and task.
Source: ThorstenMeyerAI.com
