🔍 Read the full analysis: Twice The Price, 5.7% More Intelligence 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’s Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, double Opus 4.8’s rates. Artificial Analysis scores it 5.7% higher on its Intelligence Index, while the public case for improved enterprise productivity rests on one unaudited customer story, not a controlled study.
Anthropic’s Claude Fable 5 costs twice as much per token as Opus 4.8, while scoring 5.7% higher on Artificial Analysis’s Intelligence Index, according to a July 2026 analysis by ThorstenMeyerAI.com. The report says the public evidence for a corresponding enterprise productivity gain is much thinner: it identifies one unaudited customer anecdote and no controlled human-baseline study.
Anthropic lists Fable 5 at $10 per million input tokens and $50 per million output tokens. Opus 4.8 is priced at $5 and $25, respectively; Sonnet 4.6 is listed at about $3 and $15. Those figures make Fable 5 the most expensive model in Anthropic’s active lineup, based on the source report’s review of company pricing materials.
In Artificial Analysis’s comparison, Fable 5 scored 64.9 on the Intelligence Index, against 61.4 for Opus 4.8. On GDPval-AA, a knowledge-work benchmark reported as an Elo rating, the models scored 1,932 and 1,890. The report cautions that an Elo gap should not be read as a simple percentage measure of real-world productivity; benchmark results do not establish how much faster or more effectively employees will work.
The productivity claim cited in Anthropic’s launch material comes from Stripe. According to the customer account, Fable 5 completed a migration across a 50-million-line Ruby codebase in a day, a task that otherwise would have taken a team more than two months by hand. ThorstenMeyerAI.com says it could not verify this as an audited or independently replicated measurement and classifies it as a customer testimonial. The available source material does not establish that the result generalizes to other organizations or tasks.
AI Pricing & Evidence Brief · July 2026
Twice The Price,
5.7% More Intelligence
Claude Fable 5 costs twice as much per token as Opus 4.8 and scores higher on a leading benchmark. Evidence that this translates into enterprise productivity remains limited.
A benchmark lead can justify testing a model. It cannot by itself forecast time saved, work quality, or return on investment.
01 · Price versus performance
The documented premium
Published list prices and benchmark scores are measurable. Workplace gains are a separate question—and need evidence from real tasks.
| Model | Input / 1M | Output / 1M | Intelligence Index | GDPval-AA Elo |
|---|---|---|---|---|
| Claude Fable 5 | $10 | $50 | 64.9 | 1,932 |
| Claude Opus 4.8 | $5 | $25 | 61.4 | 1,890 |
| Claude Sonnet 4.6 | ~$3 | ~$15 | Not stated | Not stated |
The reported 5.7% is the Intelligence Index score difference. Elo ratings and benchmark scores do not translate directly into a percentage increase in employee productivity.
02 · Reading the score
A lead with limits
The overall results favor Fable 5 in this comparison. Averages can conceal which workflows benefit, and by how much.
64.9 vs. 61.4
Artificial Analysis’s reported scores put Fable 5 ahead of Opus 4.8. The index measures benchmark performance, not time saved at work.
1,932 vs. 1,890
The reported Elo rating also favors Fable 5. An Elo gap is not a simple percentage measure of real-world productivity.
5 of 10 led
The source reports leading results on five of ten sub-benchmarks. A company’s value may depend on how closely its work matches those tasks.
03 · The productivity claim
One customer story
Anthropic’s launch material quotes Stripe on a large Ruby code migration. The account is illustrative, but the source review says it is unaudited and independently unreplicated.
50-million-line Ruby codebase
Stripe said Fable 5 completed a migration in a day—work that would otherwise have taken a team more than two months by hand.
Testimonial, not a trial
No controlled human-baseline study, independent audit, replication, or evidence that the result generalizes to other teams and codebases is identified.
“Fable 5 performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.”Anthropic launch post, quoting Stripe
04 · What changes the bill
List price is only a starting point
Caching, batch rates, and inference location affect costs. Actual spending depends on token mix, output volume, and the customer’s usage terms.
$7.70 blended
Source estimate per million tokens: seven cache-read tokens, two input tokens, and one output token, using a $1 cache-read rate.
$5 in · $25 out
Reported per-million-token batch prices. This option can change the bill when the workload qualifies.
1.1× multiplier
The source lists a 1.1-times price multiplier for US-only inference. Fable 5 and Mythos 5 are described as same weights, different safeguards, same pricing.
05 · A buyer’s evidence path
Test the work you actually do
The next useful evidence connects task quality and time to total cost, with a clear comparison method.
06 · Key questions
What buyers can conclude
The public evidence supports a pricing and benchmark comparison, not a quantified productivity forecast or proven return.
Is Fable 5 5.7% more productive?
No. The 5.7% refers to Intelligence Index scores, not a controlled measure of employee output.
Does the source prove it is worth the premium?
No. Whether it pays off depends on task results, quality, and total usage costs in a buyer’s workflow.
What supports the productivity claim?
A Stripe customer anecdote about a code migration. The report classifies it as unaudited and not independently replicated.
What should a buyer do next?
Compare Fable 5 with Opus 4.8 on representative tasks and track output quality alongside total token costs.
The Case for Paying More
For companies setting AI budgets, the gap between benchmark performance and workplace productivity affects whether a premium model is worth buying. A model can score higher on tests without delivering enough time savings, quality improvement or cost reduction in a particular business workflow to justify its price.
The reported benchmark results offer a reason to test Fable 5, not proof that every customer should move to it. The source analysis says the model recorded leading results on five of ten sub-benchmarks, suggesting its advantage may be more useful for some kinds of work than others. A company whose tasks match those areas could see different value from one using other workflows. The report does not provide controlled evidence to quantify that difference.
The practical issue is that the price is publicly documented, while the evidence for productivity gains is not comparably strong. Buyers relying on broad claims about employee output would need to treat the Stripe story as an example, rather than as a verified forecast for their own teams.
As an affiliate, we earn on qualifying purchases.
Price and Benchmark Comparisons
The figures in the report distinguish published rates from an estimated usage pattern. Fable 5’s list prices are $10 per million input tokens and $50 per million output tokens. ThorstenMeyerAI.com calculates a blended rate of $7.70 per million tokens under an assumed mix of seven cache-read tokens, two input tokens and one output token. That calculation uses the reported cache-read rate of $1 per million tokens; it is not a universal price for every customer’s usage.
The report also lists a batch rate of $5 per million input tokens and $25 per million output tokens, and a 1.1-times multiplier for US-only inference. These options can change the bill, but actual spending depends on how a customer uses the model. Fable 5 and Mythos 5 are described in the source as the same underlying weights with different safeguards and identical pricing.
Artificial Analysis’s Intelligence Index and GDPval-AA figures are benchmark measures, not direct measurements of time saved at work. The source report presents Fable 5 as ahead of Opus 4.8 on both, while stressing that the results do not supply a controlled comparison of human output with and without the model. Its discussion also notes that leaderboard rank and task-specific results may matter differently from an overall average.
“Fable 5 performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.”
— Anthropic’s launch post, quoting Stripe
Evidence Beyond One Customer
The available material does not identify a controlled productivity study comparing Fable 5-assisted work with a human or no-AI baseline. It also does not provide an independent audit or replication of Stripe’s account, details of the task’s evaluation method, or evidence that the reported time comparison applies to other codebases and teams.
The source report says its review examined 25 claims, but the supplied material does not include the full verification record or complete benchmark methodology. It is therefore not possible from this material alone to assess every underlying measurement. Nor does a higher score on a benchmark establish the size of any customer’s financial return.
It remains unclear how Fable 5 performs across different enterprise tasks, whether its benchmark lead produces measurable gains in day-to-day work, and how token use, caching and output volume affect real customer costs. The source places general availability around July 1, 2026; its wording is an expectation, not confirmation that availability began on that date.
What Buyers Should Measure
The next useful evidence would be customer-level evaluations that disclose the tasks tested, the comparison group, quality checks and costs, alongside independently assessed or controlled productivity studies. Those findings could show whether the benchmark advantage translates into faster or better work and for which workloads.
In the meantime, organizations considering Fable 5 can compare it with Opus 4.8 on representative tasks and track both output quality and total token costs. Such internal tests can inform a purchasing decision, but they would not substitute for a published independent study. The source materials reviewed here do not announce a specific forthcoming productivity trial or a date for additional results.
Key Questions
How much does Claude Fable 5 cost?
The source report lists $10 per million input tokens and $50 per million output tokens. Caching, batch pricing and other usage terms can affect actual costs.
Is Fable 5 5.7% more productive than Opus 4.8?
No such workplace productivity result is established by the cited figure. The 5.7% difference refers to Artificial Analysis’s Intelligence Index scores, not a controlled measure of employee output.
What evidence supports the productivity claim?
The source identifies a Stripe customer anecdote in Anthropic’s launch material about a code migration. The report says it is unaudited and lacks independent replication; it does not present it as a controlled study.
How does Fable 5 compare with Opus 4.8?
In the reported comparison, Fable 5 scored 64.9 versus 61.4 on the Intelligence Index and 1,932 versus 1,890 on GDPval-AA. Its listed input and output rates are each twice Opus 4.8’s.
Does the source establish that Fable 5 is worth the premium?
No. It documents pricing and benchmark comparisons but does not provide controlled evidence of return on investment. Whether the premium makes sense depends on a buyer’s tasks, results and usage costs.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
