TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Anthropic released Claude Fable 5.2 on November 9, 2026, and Artificial Analysis ranks it first on its Intelligence Index at Max effort. In that test, the average API cost ranges from $3.10 per task at Low effort to $26.54 at Max, while the score rises from 54 to 65. Those figures come from a benchmark and do not establish what a particular business will pay for its own work.
Anthropic released Claude Fable 5.2 on November 9, and an Artificial Analysis benchmark puts it at first place at Max effort among 74 models on the Intelligence Index. The same test shows a wide cost range: an average of $3.10 per task at Low effort and $26.54 at Max, giving businesses a benchmark-based view of what additional model effort may cost.
Artificial Analysis tested Fable 5.2 at five settings: Low, Medium, High, Xhigh and Max. Its reported index scores range from 54 at Low to 65 at Max. Medium, the default in Claude apps and Claude Code, scored 56. High and Xhigh each scored 62 as displayed by the benchmark; Max scored 65. The rankings were 15th, ninth, third, second and first, respectively, among 74 models.
The benchmark’s average API cost per task rose from $3.10 at Low to $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. Average task times also increased, from 6.5 minutes at Low to 46.4 minutes at Max. Artificial Analysis describes the cost figure as the average API cost for one task in its test, which covers nine evaluations in areas including agentic coding, business documents, science and knowledge.
Anthropic lists the same standard token prices across Fable’s effort levels: $8 per million input tokens and $40 per million output tokens. The source attributes the difference in benchmark bills to the model using more output tokens as effort increases: its reported total for the index rises from 34.9 million output tokens at Low to 275 million at Max. Anthropic says the new models produce answers up to 30% faster than version 5.1. An optional fast mode is listed at $16 per million input tokens and $80 per million output tokens, with generation speeds up to 2.5 times faster.
Claude Fable 5.2 costs by effort
Artificial Analysis benchmarks show higher effort can raise scores, task costs and completion times. These test averages are not forecasts of a business’s actual bill.
The benchmark cost ladder
Average API cost per task rises across all five effort settings.
| Effort | Score | Rank | Cost per task |
|---|---|---|---|
| Low | 54 | 15th | $3.10 |
| Medium | 56 | 9th | $3.80 |
| High | 62 | 3rd | $9.08 |
| Xhigh | 62 | 2nd | $14.83 |
| Max | 65 | 1st | $26.54 |
More effort, uneven gains
The score does not rise in step with spending, and higher effort did not win on every cited task.
- Score: 62
- Average task cost: $9.08
- FrontierCode 2.0: 49.1%
- Score: 65
- Average task cost: $26.54
- FrontierCode 2.0: 48.8%
What the prices mean
Anthropic lists the same standard token prices across effort levels; benchmark task bills differ as output token use increases.
$8 per million tokens
Standard input token price across Fable effort settings.
$40 per million tokens
Standard output token price across Fable effort settings.
Up to 2.5× faster
Optional mode costs $16 per million input tokens and $80 per million output tokens.
Choose for the workload
Use the benchmark as a comparison point, then measure your own tasks before setting a purchasing plan.
Identify the accuracy and completion time your work requires.
High scored 62 at $9.08 per benchmark task; Max scored 65 at $26.54.
Track quality, time and API bills on your prompts and task mix.
The Cost of More Model Effort
The figures offer buyers a way to compare the cost and benchmark score at each setting, while also showing that the score does not rise in step with spending. Moving from Medium to High adds six index points, from 56 to 62, and raises the reported task cost by $5.28. Xhigh costs about 63% more than High yet has the same rounded score of 62. Max adds three points over High, with an average task cost nearly three times as high.
That trade-off may matter for teams deciding which tasks warrant a longer, more expensive model run. Task time is another cost: Artificial Analysis reports an average of 17.9 minutes at High and 46.4 minutes at Max. These are benchmark averages, not a promise about response times or bills for every customer. A company’s actual results will depend on its prompts, task mix and model use.
More effort also did not improve every result cited in the source. On Artificial Analysis’s FrontierCode 2.0 coding test, Fable 5.2 scored 49.1% at High and 48.8% at Max. That example cautions against treating the Intelligence Index ranking as a guarantee that Max will perform best on a particular job.
Products Worth Considering
Samsung Galaxy Tab A11+ 8GB RAM, 256GB Storage, Gray
LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz
Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive
As an affiliate, we earn on qualifying purchases.
How Fable and Mythos Differ
Anthropic released Fable 5.2 and Mythos 5.2 as replacements for their 5.1 versions. The company describes them as the same underlying model with different access and safety controls. Fable is available to users generally and has safeguards aimed at high-risk cybersecurity work and most research biology tasks. Anthropic says a blocked query receives an answer from an older model: Opus 4.8 for cybersecurity and Opus 5 for biology.
Mythos 5.2 is restricted to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs, according to the source. Anthropic says early testing showed about 55% fewer blocks with Fable 5.2 than with Fable 5.1. These are company statements; the source does not provide independent testing of the safeguards or the reduction in blocked requests.
For a cost comparison, the source sets Fable 5.2 at High against Claude Opus 5.5 at Max. Artificial Analysis lists scores of 62 and 59, respectively, and average task costs of $9.08 and $10.04. Anthropic also says Fable 5.2 at High scored 45.3% on FrontierCode 2.0, close to Opus 5.5’s best score of 45.5%, “for roughly half the cost per task.” The benchmark figures and Anthropic’s comparison describe specific tests, not all coding work.
““for roughly half the cost per task.””
— Anthropic
What the Benchmark Cannot Show
The reported costs and scores describe Artificial Analysis’s test setup, not a typical customer’s workload. The source does not give a business-specific estimate or establish how much a given team would spend at each effort level. It also does not provide enough detail here to assess how the nine evaluations map onto a particular company’s tasks.
The source attributes claims about faster answers, safety filters and early testing to Anthropic. Those statements are not presented as independent findings. Anthropic’s accounts of prerelease accomplishments—including designing electronic devices and work on mathematical and physics problems—are likewise company claims, and do not establish what ordinary users can expect. The source material ends partway through a further benchmark comparison, so no additional result from that passage can be confirmed here.
Choosing an Effort Setting
Businesses can compare the benchmark cost ladder with the time and quality their own work requires. The source says Anthropic recommends starting with Opus 5.5 for most coding and knowledge work, then using Fable when a task calls for it. Within Fable, the benchmark suggests High may be a point to examine: it scored 62 at $9.08 per task, while Max scored 65 at $26.54. A team would need to assess those settings against its own tasks before treating the benchmark as a purchasing forecast.
Anthropic’s release and Artificial Analysis’s published figures provide the current comparison in the source, dated November 10, 2026. Customer-specific costs and results remain unknown from the information provided. The next useful evidence would be workload-based measurements showing how the effort settings affect a team’s accuracy, completion time and API bill.
Key Questions
What does Fable 5.2 cost per task?
In Artificial Analysis’s Intelligence Index test, reported average API cost ranges from $3.10 at Low effort to $26.54 at Max. These are benchmark task costs, not fixed prices for every user request.
Which effort level ranked first?
Max effort ranked first among 74 models on the reported Intelligence Index, with a score of 65. High and Xhigh scored 62 as displayed by the benchmark.
Are token prices different at each effort level?
Anthropic lists standard prices of $8 per million input tokens and $40 per million output tokens for all five levels. The benchmark’s task costs differ as output token use rises with effort.
Who can use Mythos 5.2?
According to the source, Mythos 5.2 is limited to vetted organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Fable 5.2 is the version described as available to all users.
Does Max perform best on every task?
The source does not establish that. In the cited FrontierCode 2.0 test, Fable scored 49.1% at High and 48.8% at Max, illustrating that higher effort did not produce the higher score in that result.
Source: Thorsten Meyer AI
