🔍 Read the full analysis: Claude Fable 5.2 For AI Tasks: Comparing The Cost Of Each Effort Setting 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
Artificial Analysis reports that Claude Fable 5.2 scored from 54 to 65 on its Intelligence Index as effort increased from Low to Max, while estimated cost per task rose from $3.10 to $26.54. High scored 62 for $9.08, compared with 65 for $26.54 at Max; these figures reflect the index test, not a universal cost for every task.
Claude Fable 5.2 scored 54 to 65 on the Artificial Analysis Intelligence Index across five effort settings, while the estimated cost of an index task rose from $3.10 at Low to $26.54 at Max. The results, accessed November 10, show why businesses choosing the model may need to weigh incremental benchmark gains against higher bills and longer run times.
Artificial Analysis tested Fable 5.2 at Low, Medium, High, Xhigh and Max on an index made up of nine evaluations covering agentic coding, business documents, science and knowledge. Its reported average cost per task increased with effort: $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max. Scores were 54, 56, 62, 62 and 65, respectively. The figures are specific to this index and its API-cost estimates, not a fixed charge for every use.
The step from Medium to High added six points for $5.28 more per task. Xhigh cost about 63% more than High but had the same rounded score, 62. Max added three points over High while costing nearly three times as much. Average task time also rose, from 6.5 minutes at Low to 46.4 minutes at Max.
Artificial Analysis reports that each effort level has the same listed token prices: $8 per million input tokens and $40 per million output tokens. The difference in estimated task cost reflects how much the model generates while reasoning. In the index run, reported output tokens rose from 34.9 million at Low to 275 million at Max. These are totals for the test set, not tokens used in a single task.
Claude Fable 5.2 The cost of each effort setting
Artificial Analysis scores rose with effort, but so did estimated task cost and time. The figures describe one benchmark index—not a guaranteed price or result for every workload.
Five settings, rising costs
Artificial Analysis estimated cost per task for its index test; scores are shown alongside each estimate.
| Effort | Score | Estimated cost |
|---|---|---|
| Low | 54 | $3.10 |
| Medium | 56 | $3.80 |
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
What the added effort buys
The benchmark shows gains are not uniform across settings or tests.
A notable score increase
Moving from Medium to High added six index points for $5.28 more per estimated task.
Same rounded score
Xhigh and High both scored 62, while Xhigh cost about 63% more.
Highest score, highest cost
Max added three points over High while costing nearly three times as much.
Why effort changes the bill
Listed token prices stay the same across effort levels; estimated test costs rise as output grows.
- $8 per million input tokens
- $40 per million output tokens
- 34.9 million at Low
- 275 million at Max
Test before choosing
A benchmark comparison can guide a trial, but it cannot determine the best setting for every business task.
Use representative tasks from your actual workflow.
Track output quality, completion time and actual API usage.
Check whether any improvement justifies the added cost and wait time.
The Cost of More Model Effort
The results give organizations a concrete way to think about quality, latency and spending before assigning a model setting to routine work. A higher effort level may be justified for difficult tasks where a better result is valuable, but the index does not show that every request benefits from paying for Max. The same benchmark recorded a 62 score at High and Xhigh, despite Xhigh’s higher estimated cost and longer average runtime.
There is also a limit to what a single aggregate score can establish. Artificial Analysis’s reported FrontierCode 2.0 coding result for Fable 5.2 was 49.1% at High and 48.8% at Max. That one test suggests that extra effort did not improve every measured result. Teams considering deployment would need to test their own workloads and compare output quality, completion time and usage costs, rather than assume that the highest setting is best.
Products Worth Considering
LG UltraGear 25G590B-B 25-Inch Full HD 1000Hz Gaming Monitor
LG UltraGear 32GX850A-B 32-inch 4K UHD OLED Gaming Monitor
Sonos Arc Ultra Soundbar with Dolby Atmos, Black
As an affiliate, we earn on qualifying purchases.
Fable 5.2 and Its Effort Dial
Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, describing them as replacements for the 5.1 versions. The supplied source says Fable is available to users with safeguards for high-risk cybersecurity and research-biology requests. Mythos is offered to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. These availability and safety details are Anthropic’s descriptions, not findings from the Artificial Analysis cost test.
Fable 5.2 provides five effort options, with Medium as the default in Claude apps and Claude Code, according to the source. Anthropic lists standard prices of $8 per million input tokens and $40 per million output tokens. It also offers an optional fast mode at higher listed prices. The source says Anthropic recommends beginning with Claude Opus 5.5 for most coding and knowledge work, and moving to Fable when a task calls for it.
Artificial Analysis places Fable 5.2 at #1 of 74 models at Max, with a score of 65. Its reported scores put the same model at #3 at High and #2 at Xhigh. Separately, the source compares Fable at High with Opus 5.5 at Max: 62 for an estimated $9.08 per task versus 59 for $10.04. That comparison is also tied to the index and should not be read as a guarantee for every application.
“For most coding and knowledge work, start with Claude Opus 5.5 and move up to Fable only when the task needs it.”
— Anthropic, as quoted in the supplied source
What the Benchmark Cannot Settle
The reported costs and scores describe Artificial Analysis’s particular index test. The supplied material does not provide enough detail to establish how closely its average task cost matches a specific business’s prompts, agent setup or usage pattern. Actual bills can depend on those factors, including how much output a model produces. The index figures should not be treated as price quotes for other workloads.
The source also gives several claims from Anthropic’s prerelease testing, including complex scientific and engineering achievements. Those accounts are vendor claims and are not independently verified here; they do not establish what typical users will achieve. The material does not include independent testing of the claimed speed improvement or a full breakdown of performance by individual index evaluation. It also does not establish whether effort settings produce consistent gains across repeated runs.
It remains unclear how the reported average time and cost will translate to live use, or how widely organizations will be able to access Mythos under Anthropic’s verification programs. The benchmark ranks and prices are a snapshot dated November 10, 2026 and may change as models, pricing and tests are updated.
Testing Settings on Real Work
For organizations deciding whether to use Fable 5.2, the next step is to compare settings on representative tasks, recording quality, completion time and actual API usage. A practical test could begin at Medium or High, then check whether higher effort improves results enough to justify its added cost and wait time. The supplied source does not give a date for further Artificial Analysis testing or a new Anthropic release.
Readers should also watch for updated benchmark results, clearer workload-specific cost data and independent verification of Anthropic’s performance claims. Until then, the published index offers a useful comparison of five settings, but it does not identify one setting as the right choice for every user.
Key Questions
How much does a Fable 5.2 task cost at each effort level?
Artificial Analysis estimates $3.10 at Low, $3.80 at Medium, $9.08 at High, $14.83 at Xhigh and $26.54 at Max for a task in its Intelligence Index. These are benchmark-specific estimates, not guaranteed costs for other work.
Which effort level had the highest index score?
Max scored 65, the highest reported Fable 5.2 result in the comparison. It also had the highest estimated cost and longest average task time: $26.54 and 46.4 minutes.
Does a higher effort setting always produce a better result?
No. In the supplied results, High and Xhigh both scored 62 when rounded, while Fable 5.2 scored slightly higher on the reported FrontierCode 2.0 test at High than at Max. Those results do not establish which setting will work best on a particular user’s tasks.
Why do the settings cost different amounts if token prices are the same?
The supplied source says list prices are the same across Fable effort levels, while output-token use increases at higher effort. Artificial Analysis reported 34.9 million output tokens across its index at Low and 275 million at Max, contributing to the difference in estimated task costs.
Source: ThorstenMeyerAI.com
