AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What Does It Cost To Put Claude Fable 5.2 To Work? on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

Anthropic released Claude Fable 5.2 on Nov. 9, 2026, with five effort settings that vary the amount of model reasoning and the resulting benchmark cost. Artificial Analysis reported costs ranging from $3.10 per Intelligence Index task at Low effort to $26.54 at Max; the top score was 65, while High scored 62 for $9.08.

Anthropic released Claude Fable 5.2 on November 9, offering five effort settings whose benchmark task costs range from $3.10 to $26.54, according to Artificial Analysis. On the index, Fable scored 54 at Low effort and 65 at Max, a trade-off businesses face when deciding how much reasoning time and expense a task warrants, as explored in analysis of Fable 5.1’s cost factors.

Artificial Analysis tested Fable 5.2 on its Intelligence Index, which includes nine evaluations covering agentic coding, business documents, science and knowledge. Its reported “cost per task” is the average API cost for one task in that test, not a fixed price for every real-world request. At the default Medium setting, the model scored 56 and cost $3.80 per task. High scored 62 at $9.08; Xhigh also scored 62, at $14.83; and Max scored 65 at $26.54.

The costs reflect usage rather than different token rates across effort settings, a question also raised by scrutiny of AI flat rates. Anthropic lists Fable 5.2 at $8 per million input tokens and $40 per million output tokens. Artificial Analysis reported that the model used 34.9 million output tokens across its index at Low effort and 275 million at Max. Average task time also rose, from 6.5 minutes at Low to 46.4 minutes at Max.

Anthropic says Fable 5.2 replaces Fable 5.1 and produces answers up to 30% faster than its predecessor, following Anthropic’s Claude verification work. It also offers an optional fast mode, which Anthropic says generates tokens up to 2.5 times faster, at rates of $16 per million input tokens and $80 per million output tokens. These are company statements; the supplied material does not include independent verification of the speed claims.

At a glance
reportWhen: Released November 9, 2026; benchmark fi…
The developmentAnthropic released Claude Fable 5.2 with five selectable effort levels, and Artificial Analysis published benchmark scores and per-task costs for each.
AI Costs · Briefing

What It Costs to Put Claude Fable 5.2 to Work

Five effort settings trade benchmark performance against API usage and waiting time. The best fit depends on the task—not simply the highest score.

$26.54
Max average cost per benchmark task
65
Max Intelligence Index score
$8 / $40
Per million input / output tokens
Nov 9, 2026
Release date

Effort means different benchmark costs

Artificial Analysis reports average cost per Intelligence Index task; these are not fixed prices for individual requests.

EffortScoreCost per taskAverage time
Low54$3.106.5 minutes
Medium56$3.80Not reported
High62$9.0817.9 minutes
Xhigh62$14.83Not reported
Max65$26.5446.4 minutes

What the trade-off looks like

More effort can raise the score, but gains and costs vary by setting and task.

High
  • Score: 62
  • Average cost: $9.08 per task
  • Average time: 17.9 minutes
Max
  • Score: 65 on the Intelligence Index
  • Average cost: $26.54 per task
  • Average time: 46.4 minutes

Choose the model and access for the work

Anthropic recommends starting with Opus 5.5 for most coding and knowledge work, then using Fable when a task calls for it.

General access

Claude Fable 5.2

Offers five effort settings. Anthropic says safeguards block certain high-risk cybersecurity work and most research biology tasks.

Restricted access

Claude Mythos 5.2

Limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs.

Benchmark results need real-world checks

Index averages may not match a company’s prompts, output volume, task mix, costs, or response times.

1
Test representative work

Run tasks that reflect your actual workload at several effort settings.

2
Track quality and usage

Compare results alongside API costs and wait times.

3
Choose per task

Use higher effort where its improvement is worth the added cost and time.

Benchmark caveatReported costs are Artificial Analysis averages, not quotes for individual prompts. Anthropic’s speed and safety-block claims are company-reported and were not independently verified in the supplied material.
The takeaway
Start with the least effort that meets the task’s needs, then validate the quality, cost, and waiting time on your own work.

The Price of More Reasoning

The figures make the effort setting a budget decision, not just a quality setting. Moving from Medium to High adds six points on the reported index for $5.28 more per task. Going from High to Xhigh costs substantially more, while both settings receive the same rounded score of 62. Max adds three points over High but costs nearly three times as much and takes much longer on average.

For organizations running many model tasks, small differences in average API cost can add up. But the index figures are not a forecast of a particular company’s bill: actual spending depends on prompts, output volume, task mix and usage. Speed also matters where users are waiting for a result. The benchmark’s 46.4-minute average at Max may be a poor fit for workflows that need quick responses, even if the highest index score is useful for some demanding work.

Artificial Analysis also reported that, on its FrontierCode 2.0 agentic coding test, Fable 5.2 at High scored 49.1%, slightly above its 48.8% result at Max. That result cautions against assuming that the most expensive setting will perform best on every task.

Vitamix 5-Speed Immersion Blender, 625-Watt, Stainless Steel/Black

Vitamix 5-Speed Immersion Blender, 625-Watt, Stainless Steel/Black

As an affiliate, we earn on qualifying purchases.

Fable, Mythos and Access

Anthropic introduced two names for the release: Fable 5.2 for general users and Mythos 5.2 for organizations accepted into its Cyber or Life Sciences Verification Programs. The source describes Mythos as the same underlying model without some of Fable’s safety restrictions. Anthropic says Fable’s safeguards block certain high-risk cybersecurity work and most research biology tasks; blocked queries are answered by an older model instead. Mythos is restricted to vetted organizations working in those dual-use fields.

Anthropic says early testing found about 55% fewer blocks than with Fable 5.1. That is a company-reported result, and the supplied material does not describe the testing method. The company also recommends that most users begin with Claude Opus 5.5 for coding and knowledge work, and move to Fable when a task calls for it.

Artificial Analysis ranked Fable 5.2 at Max first among 74 models on its Intelligence Index as of November 10. The next two listed positions were also Fable at Xhigh and High, each with a rounded score of 62. In the same index, the source reports Opus 5.5 at Max scoring 59 for $10.04 per task, compared with Fable at High scoring 62 for $9.08. These figures come from the named benchmark, not a guarantee of performance or savings on other work.

“For most coding and knowledge work, start with Claude Opus 5.5 and move up to Fable only when the task needs it.”

— Anthropic

What the Benchmark Cannot Show

The reported costs are averages for Artificial Analysis’s Intelligence Index tasks. The source does not establish how closely those tasks match a specific company’s workload, or what a particular user’s bill will be. It also does not provide a detailed breakdown of individual task costs, so the average should not be treated as a quote for each prompt.

Anthropic’s claims about faster answers, fewer safety blocks and prerelease demonstrations are company-reported. The source says the demonstrations—including a multi-agent proof effort—were not independently verified. It remains unclear how consistently Fable’s benchmark advantage will translate to routine work, how frequently users will need High or Max effort, and what latency they will experience in their own setups.

Measure Costs on Real Work

For teams considering Fable 5.2, the next practical step is to test representative tasks at several effort settings and record both output quality and usage costs. The published results suggest that Medium or High may be a more economical starting point than Max, but the right setting depends on the work and the value of any improvement.

Further information that could clarify the trade-off includes independent results on additional task types, per-task cost distributions, and measured response times in production. Anthropic’s stated guidance is to begin with Opus 5.5 for most coding and knowledge work, then use Fable when needed. No future benchmark release or additional pricing change is specified in the source material.

Key Questions

How much does Claude Fable 5.2 cost per benchmark task?

Artificial Analysis reported an average of $3.10 at Low effort, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max on its Intelligence Index. These are benchmark averages, not guaranteed prices for individual requests.

What does the Max setting add?

On the Artificial Analysis Intelligence Index, Max scored 65, compared with 62 at High, and averaged 46.4 minutes per task versus 17.9 minutes at High. Its reported average cost was $26.54 per task, against $9.08 at High.

Does every effort setting have a different token price?

No. Anthropic lists the same standard rates for Fable 5.2 across settings: $8 per million input tokens and $40 per million output tokens. The benchmark costs differ as the amount of model output and reasoning changes.

Who can use Claude Mythos 5.2?

According to the source, Mythos 5.2 is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Fable 5.2 is the version made available to general users.

Is Max always the best setting?

No. Max had the highest score on this index, but the supplied results show High scored slightly better than Max on FrontierCode 2.0. The best setting depends on the task, cost and acceptable wait time.

Source: ThorstenMeyerAI.com

You May Also Like

Technology Operations Signal Monitor: The Future Of Flipper Zero Development

A new signal monitor tracks updates on Flipper Zero’s development, offering role-specific alerts for small software teams to stay ahead.

Apple Wants Blacklisted Chinese RAM — and That Tells You How Bad the Squeeze Got

Apple is lobbying the Trump administration to use CXMT memory chips as rising DRAM prices squeeze Macs and iPads, the Financial Times reported.

732 Bytes to Root. One Hour of Scan Time.

A new Linux kernel privilege escalation bug was found using AI, requiring only 732 bytes and one hour of scan time, collapsing security costs.

Why External GPU Docks Appeal to Performance-Hungry Creators

Discover the top external GPU docks for 2026, including options for high performance, value, and ease of use. Find the perfect eGPU dock for your setup today.