🔍 Read the full analysis: Claude Fable 5.2 Tops The AI Model Rankings: The Cost Of Each Task on ThorstenMeyerAI.com
Get smart everyday buys 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 Nov. 9, and Artificial Analysis ranked its Max-effort setting first among 74 models on its Intelligence Index the next day. The score was 65, but an index task averaged $26.54 at Max, compared with $3.10 at Low; the ranking is a benchmark result, not a guarantee of performance on every task.
Anthropic released Claude Fable 5.2 on Nov. 9, and by the following day the model ranked No. 1 among 74 models on Artificial Analysis’s Intelligence Index when run at Max effort. That result averaged $26.54 per benchmark task; the same model at Low effort cost $3.10 per task, showing how much its ranking and bill depend on the selected setting.
Artificial Analysis scores Fable 5.2 at 65 points at Max effort. Its High and Xhigh settings each scored 62, ranking third and second respectively, while Medium scored 56 and Low scored 54. These are scores on the index, which covers nine evaluations spanning areas such as agentic coding, business documents, science and knowledge; they do not establish that Fable is best for every use or workload.
The published average 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 time also increased, from 6.5 minutes at Low to 46.4 minutes at Max. Artificial Analysis attributes the differences to how much the model reasons at each effort level: its reported output-token total for the full index increased from 34.9 million at Low to 275 million at Max.
Anthropic lists the same standard API prices across the settings: $8 per million input tokens and $40 per million output tokens. The optional fast mode is priced at $16 and $80 per million tokens, respectively, and Anthropic says it generates tokens up to 2.5 times faster. The company also says Fable 5.2 writes answers up to 30% faster than its 5.1 predecessor; those are vendor-reported claims, distinct from the benchmark’s average task times.
Claude Fable 5.2 tops the rankings at a steep task cost
Artificial Analysis ranked Fable 5.2 first at Max effort, but its benchmark results show that the highest score comes with higher cost and longer task time.
Effort changes the bill
The benchmark’s average cost rose with each effort setting.
Score versus spend
High scored close to Max on the index at a lower average task cost.
| Effort | Index score | Avg. task cost |
|---|---|---|
| High | 62 | $9.08 |
| Xhigh | 62 | $14.83 |
| Max | 65 | $26.54 |
Results vary by task
More effort did not improve every measured outcome.
- Max scored 65.
- High and Xhigh each scored 62.
- High scored 49.1%.
- Max scored 48.8%.
A practical evaluation
Teams can compare settings on representative work and track quality alongside cost.
Use examples that reflect your team’s work.
Test High against Max, then assess quality, cost and time.
Use the results to select a setting for each kind of task.
The Price of Max-Effort Results
The ranking gives teams a more useful decision than a single winner: which effort setting earns its extra cost for a particular job? On this index, moving from Medium to High raised the score from 56 to 62 for $5.28 more per task. Moving from High to Xhigh raised the listed average cost to $14.83 while the rounded score stayed at 62. Max scored three points above High, but cost nearly three times as much and took longer on average.
Those trade-offs matter for organizations running many model calls or waiting on responses in interactive workflows. A benchmark’s average cost is not a quote for every real-world task, and a higher index score alone does not settle whether a slower, more expensive setting is worthwhile. Artificial Analysis also reported that on its FrontierCode 2.0 coding evaluation, Fable scored 49.1% at High and 48.8% at Max. The result is a reminder that more effort did not improve every measured outcome.
There is also an internal comparison for buyers considering Anthropic’s other model. Artificial Analysis lists Opus 5.5 at Max at 59 points and $10.04 per task, compared with Fable 5.2 at High at 62 points and $9.08. That may help frame a choice, but it is still a comparison on this index, not evidence that one model is cheaper or stronger across every use case.
Products Worth Considering
Dell UltraSharp U2724DE 27-Inch WQHD IPS Black Monitor
Dell UltraSharp U2424H 24-Inch FHD Monitor
LG UltraGear 27-inch QHD OLED Gaming Monitor, 240Hz
As an affiliate, we earn on qualifying purchases.
Two Names, Different Safety Access
Anthropic released Fable 5.2 and Mythos 5.2 as replacements for the 5.1 versions. The source describes them as the same underlying model with different safety controls and access: Fable is available to users generally, while Mythos is restricted to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs. Anthropic says Fable’s safeguards block high-risk cyber requests and most research-biology tasks; when a request is blocked, the company says an older model handles it instead.
Anthropic says early testing found about 55% fewer blocks than with Fable 5.1. The company also publicized high-end prerelease examples, including designing electronic devices and work on mathematical and physics problems. Those examples are Anthropic’s accounts, not independent evaluations of typical customer results. The company reported one Fermat’s Last Theorem formalization run using about 1,000 agents: Fable had proved 86% of weighted statements after 18 hours and all after 30 hours, while Opus 5.5 was at 26% and 27% at those checkpoints. The source says these were single runs and vendor figures, not independently verified results.
For its own deployment guidance, Anthropic recommends starting with Opus 5.5 for most coding and knowledge work and moving to Fable when a task calls for it. Medium is the default effort setting in Claude apps and Claude Code, according to the source material. That advice underscores that the benchmark’s top-ranked configuration is not necessarily the company’s recommended default for routine work.
“The new models write their answers up to 30% faster.”
— Anthropic, on Fable 5.2 compared with the 5.1 models
Limits of the Ranking and Cost Data
The figures supplied here do not establish how representative the Intelligence Index is of any particular company’s workload. The $3.10 to $26.54 costs are benchmark averages for one task in Artificial Analysis’s test, not fixed prices for every prompt or project. The ranking also depends on effort level, and scores are shown rounded, so equal displayed scores do not necessarily mean identical underlying results.
It remains unclear how Fable and Mythos perform across broader customer use, how often organizations will need the higher settings, and how Mythos access will work for applicants beyond Anthropic’s description of its verification programs. The source material gives no independent verification of the company’s safety, speed or prerelease achievement claims. It also does not provide enough information to assess whether the model’s longer benchmark times translate into similar delays for other tasks.
How Teams May Test Fable
Businesses evaluating Fable can compare effort settings on their own representative tasks, tracking both output quality and total cost rather than relying on the index score alone. The published benchmark suggests that High may be a practical point to test against Max: it scored 62 for $9.08 per task, while Max scored 65 for $26.54. That is a benchmark-based comparison, not a recommendation for every workload.
Further independent evaluations and customer results may clarify whether the index’s performance differences carry over to production. Organizations seeking Mythos will also need to follow Anthropic’s verification process for the Cyber or Life Sciences programs. No additional release date or next benchmark milestone is specified in the source material.
Key Questions
What happened with Claude Fable 5.2?
Anthropic released Fable 5.2 on Nov. 9, 2026. Artificial Analysis ranked its Max-effort setting first among 74 models on its Intelligence Index the next day, with a score of 65.
How much does a Fable 5.2 task cost?
Artificial Analysis reported average index-task costs from $3.10 at Low effort to $26.54 at Max. These are averages for tasks in that benchmark, not guaranteed prices for other workloads.
Does Max effort always perform better than High?
No. Max scored 65 versus High’s 62 on the Intelligence Index, but on Artificial Analysis’s FrontierCode 2.0 test, High scored 49.1% and Max 48.8%. Results vary by evaluation.
What is the difference between Fable 5.2 and Mythos 5.2?
The source describes them as the same underlying model with different safety controls and access. Fable is available to users generally; Mythos is limited to organizations accepted into Anthropic’s Cyber or Life Sciences Verification Programs.
Are Anthropic’s reported prerelease achievements independently verified?
Not in the material provided. The mathematical, physics and device-design examples are Anthropic’s own accounts; the source says the Fermat run figures were single-run vendor data, not independently verified.
Source: ThorstenMeyerAI.com
