🔍 Read the full analysis: Pick An Effort Level: What Claude Fable 5.2 Costs For Each Task on ThorstenMeyerAI.com
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
TL;DR
Anthropic released Claude Fable 5.2 on November 9, 2026, with five effort settings whose Artificial Analysis benchmark costs range from $3.10 to $26.54 per task. The benchmark score rose from 54 at Low to 65 at Max, but Max took an average of 46.4 minutes per task; the results are test-specific, not a forecast of every customer’s bill.
Anthropic released Claude Fable 5.2 on November 9, and benchmark figures published by Artificial Analysis show that its five effort settings cost an average of $3.10 to $26.54 per task on the group’s Intelligence Index. Scores ranged from 54 at Low to 65 at Max, while Max took an average of 46.4 minutes per task, giving businesses a concrete trade-off between benchmark performance, cost and wait time.
Artificial Analysis tested Fable 5.2 at Low, Medium, High, Xhigh and Max on an index spanning nine evaluations in areas including agentic coding, business documents, science and knowledge. Its reported average cost per task was $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. The corresponding displayed scores were 54, 56, 62, 62 and 65.
The time reported per task also increased 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 ranked the settings 15th, ninth, third, second and first, respectively, among 74 models or configurations. These costs are averages on that benchmark, not guaranteed charges for a particular customer request.
The largest score increase in the listed settings came between Medium and High: six points for $5.28 more per benchmark task. Xhigh cost 63% more than High while both displayed a score of 62. Max added three points over High and cost nearly three times as much. Scores are shown as rounded by Artificial Analysis, so equal displayed scores do not establish identical underlying results.
Pick an effort level What Fable 5.2 costs per task
Artificial Analysis benchmark results show a trade-off between score, task cost and wait time across five Claude Fable 5.2 effort settings. These are test averages, not forecasts of customer bills.
Effort setting trade-offs
Higher settings generally cost more and take longer in this benchmark; High and Xhigh share the same displayed score.
| Setting | Cost per task | Score | Time | Rank |
|---|---|---|---|---|
| Low | $3.10 | 54 | 6.5 min | 15th |
| Medium | $3.80 | 56 | 8.3 min | 9th |
| High | $9.08 | 62 | 17.9 min | 3rd |
| Xhigh | $14.83 | 62 | 26.4 min | 2nd |
| Max | $26.54 | 65 | 46.4 min | 1st |
What the extra effort buys
The largest score increase among the listed settings comes from Medium to High.
The score rises six points, from 56 to 62, for $5.28 more per benchmark task.
The displayed score stays at 62 while cost rises 63% and average time increases.
Max adds three displayed points, costs nearly three times as much, and averages 46.4 minutes per task.
Choose for the workload
Benchmark rankings are a starting point for testing, not a universal setting guide.
- Low and Medium had lower benchmark costs and shorter average times.
- Consider testing them on routine requests where speed and spend matter.
- High, Xhigh and Max scored higher on the overall index.
- Test against representative tasks when extra benchmark performance may justify more time and cost.
Benchmark results have limits
Artificial Analysis tested an index spanning nine evaluations, including agentic coding, business documents, science and knowledge.
Published API rates
Anthropic lists separate token prices for standard and optional fast mode.
Input and output
$8 per million input tokens and $40 per million output tokens.
Higher token rates
$16 per million input tokens and $80 per million output tokens; Anthropic says it generates tokens up to 2.5 times faster.
Fable and Mythos
Fable is generally available. Mythos is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs.
The Cost of More Model Effort
The figures make the effort setting a practical spending and latency choice, rather than a simple quality toggle. A team handling routine requests may value lower cost and faster responses; a job where a few additional benchmark points matter may justify a more expensive setting. For interactive work, a reported average of nearly 47 minutes at Max may be a bigger constraint than the task’s benchmark cost.
The results also caution against assuming that the highest setting always performs best. Artificial Analysis reported that Fable 5.2 at High scored 49.1% on FrontierCode 2.0, slightly above its 48.8% at Max. That result applies to this particular agentic-coding test; it does not show that High will outperform Max across other tasks.
Artificial Analysis also compared Fable at High with Anthropic’s Claude Opus 5.5 at Max: Fable scored 62 for $9.08 per task, against Opus 5.5’s score of 59 for $10.04. This is a comparison within the cited index, not evidence that one option will be better for every workload or usage pattern.
Products Worth Considering
LG UltraGear 32GX850A-B 32-inch 4K UHD OLED Gaming Monitor
LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz
DJI Mini 4K Drone with 4K UHD Camera
As an affiliate, we earn on qualifying purchases.
Release, Pricing and Safety Tiers
Fable 5.2 replaces Fable 5.1. Anthropic also announced Claude Mythos 5.2, which the source describes as the same underlying model with fewer safety restrictions, available only to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Fable is available to users generally and includes safeguards intended to block certain high-risk cyber and biology requests.
Anthropic lists standard API prices of $8 per million input tokens and $40 per million output tokens for Fable 5.2. The company says its optional fast mode generates tokens up to 2.5 times faster, at $16 per million input tokens and $80 per million output tokens. The Artificial Analysis task-cost figures reflect benchmark usage and model reasoning consumption, rather than a separate per-task price set by Anthropic.
According to Anthropic, Fable 5.2 answers up to 30% faster than Fable 5.1 and early testing produced about 55% fewer safety blocks. Anthropic recommends starting with Claude Opus 5.5 for most coding and knowledge work, then moving to Fable when a task calls for it. Those are the company’s statements and guidance, not independent findings about all customer workloads.
“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 Results Have Limits
The figures describe Artificial Analysis’s nine-evaluation index and its average task costs. The source does not establish how closely those evaluations match any specific company’s workload, how often customers will select each effort setting, or what their actual bills will be. Real costs can vary with prompt size, output length and the amount of reasoning tokens used.
The source also reports Anthropic’s accounts of prerelease demonstrations, including work on electronic devices, black-hole physics and a large agent-based proof effort. Those are vendor-reported results, not independently verified demonstrations. The supplied material does not include a response from Artificial Analysis beyond its published benchmark data or independent replication of Anthropic’s claims.
It remains unclear how the benchmark rankings and prices will compare with broader production use, and whether the displayed score differences will translate into meaningful improvements for particular tasks. The reported figures are dated November 10, 2026, and may change as tests or model offerings are updated.
Test Settings on Real Workloads
Organizations considering Fable 5.2 can compare its effort settings on representative tasks, recording quality, completion time and token costs rather than relying on a single index score. The benchmark suggests High as a point to test against Max when a task needs more effort than the default, but it does not prescribe a setting for every use case.
Customers should also check Anthropic’s current API pricing and availability before deployment, particularly if they plan to use fast mode or require access to Mythos. Further independently reported benchmark results or production cost data would help clarify whether the index trade-offs hold across different workloads.
Key Questions
How much does a Fable 5.2 task cost?
In Artificial Analysis’s Intelligence Index test, average cost ranged from $3.10 at Low to $26.54 at Max. These are benchmark averages, not fixed charges for all customer tasks.
Which effort setting had the highest reported score?
Max scored 65 on the cited index, ranked first among 74 entries in the source. It averaged 46.4 minutes and $26.54 per task in that test.
Is Max always better than High?
No. Max scored higher on the overall index, but the source says High scored 49.1% on FrontierCode 2.0, compared with 48.8% for Max. Results depend on the evaluation and task.
What are Anthropic’s standard API rates for Fable 5.2?
The source lists $8 per million input tokens and $40 per million output tokens. Optional fast mode is listed at $16 and $80 per million, respectively.
Who can use Claude Mythos 5.2?
According to the source, Mythos is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Fable 5.2 is the version offered to users generally, with safety safeguards.
Source: ThorstenMeyerAI.com
