🔍 Read the full analysis: The Numbers Behind Double-Priced AI And A 5.7% Gain 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 reports a 5.7% higher Intelligence Index score for Fable 5, but the cited enterprise productivity case is a single customer testimonial, not an audited or independently replicated study.
Anthropic’s Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens, twice the listed rates for Opus 4.8, while its score on Artificial Analysis’s Intelligence Index is 5.7% higher. A July 2026 report from ThorstenMeyerAI.com says the public case that Fable 5 improves enterprise productivity is much less developed: it identifies one customer anecdote and no controlled, audited productivity study.
The pricing figures are attributed to Anthropic’s launch announcement, platform pricing documentation and product page. The report lists Opus 4.8 at $5 per million input tokens and $25 per million output tokens, and Sonnet 4.6 at about $3 and $15, respectively. On those figures, Fable 5 is the most expensive model in Anthropic’s active lineup described by the report. A lower-cost batch tier and prompt-caching discounts can reduce effective spending; the article estimates a blended rate of $7.70 per million tokens using a stated mix of cached reads, input and output.
On the benchmark comparison cited, Fable 5 scored 64.9 on the Intelligence Index, against 61.4 for Opus 4.8. Artificial Analysis’s GDPval-AA knowledge-work Elo scores were 1,932 and 1,890, respectively. These results describe performance on benchmarks, not measured gains in workers’ productivity. The source also says Fable 5 set records in five of the ten sub-benchmarks, suggesting that results may differ by task; it does not establish that a particular buyer’s workload will benefit enough to offset the higher price.
The report identifies one public human-baseline productivity example: Anthropic said a customer, Stripe, used Fable 5 to complete a migration across a 50-million-line Ruby codebase in a day, a task the company said would otherwise have taken a team more than two months by hand. That is a customer story reported by Anthropic. The report says it found no published methodology, independent audit or replication supporting it as a measured productivity result.
AI Economics · July 2026 Report
The Numbers Behind Double-Priced AI And A 5.7% Gain
Claude Fable 5 lists at twice Opus 4.8’s price and leads its Intelligence Index score by 5.7%. The public productivity case, however, rests on one customer testimonial without a published independent audit or replication.
The economics are fully documented. The productivity story is not.
ThorstenMeyerAI.com report
01 / Price comparison
A clear premium, with discounts in the mix
The source report lists Fable 5 at $10 per million input tokens and $50 per million output tokens. Opus 4.8 is half that rate; Sonnet 4.6 is lower still. Prompt caching and batch processing can reduce effective spend. The report estimates a blended rate of $7.70 per million tokens for a stated usage mix.
Claude Fable 5
Highest listed rate in report
Claude Opus 4.8
Half the Fable 5 list price
Claude Sonnet 4.6
Lower-cost listed alternative
02 / Benchmark comparison
Higher scores are not productivity measurements
Artificial Analysis figures cited in the report show a 5.7% lead on the Intelligence Index and a smaller raw difference on GDPval-AA. Benchmarks describe performance on defined tests; they do not establish time saved across a company’s day-to-day work.
| Measure | Fable 5 | Opus 4.8 | What it indicates |
|---|---|---|---|
| Intelligence Index | 64.9 | 61.4 | Aggregate benchmark score |
| GDPval-AA Elo | 1,932 | 1,890 | Knowledge-work benchmark rating |
| Input price / million | $10 | $5 | Published list rate cited |
| Output price / million | $50 | $25 | Published list rate cited |
03 / Evidence check
One striking account; broad returns remain unverified
The report identifies a single public enterprise example. It says the account lacks published methods, an independent audit, and replication, so readers cannot assess how the comparison was calculated or how widely it applies.
Stripe code migration
Anthropic said Stripe used Fable 5 to complete a migration across a 50-million-line Ruby codebase in one day, a task it said would otherwise have taken a team more than two months by hand.
Not an audited productivity study
The report found no published controlled methodology, independent verification, or replication supporting the story as a measured productivity result. One account cannot establish typical enterprise gains.
04 / Buyer’s decision path
Measure the work that matters to you
A benchmark lead may matter when your tasks resemble the areas where Fable 5 performs strongly. A workload-specific pilot can show whether that advantage offsets the higher cost.
Choose tasks
Use representative work from your own team and workflows.
Set a baseline
Define human or no-AI time, quality, and completion criteria.
Run a pilot
Compare models on the same tasks and track token usage.
Calculate return
Weigh measured time and quality against total effective cost.
At a glance / Benchmark scores
Index lead: 3.5 points
The score difference equals a 5.7% relative lead over Opus 4.8 on the cited Intelligence Index. The GDPval-AA gap is 42 Elo points; Elo ratings are logarithmic, so a raw percentage does not directly describe head-to-head performance.
Relative bar lengths use the Fable 5 score as the 100% reference. Benchmark results do not directly measure workplace productivity.
Key questions
What buyers can conclude today
How much does Fable 5 cost?
The report lists $10 per million input tokens and $50 per million output tokens. Cache reads and batch processing may lower effective cost.
How does it compare with Opus 4.8?
It costs twice as much and scores 64.9 versus 61.4 on the Intelligence Index. GDPval-AA scores are 1,932 versus 1,890.
Is higher enterprise productivity proven?
Not by the cited public evidence. The report identifies one Anthropic customer story and no published controlled study, independent audit, or replication.
Could the premium still pay off?
Possibly, depending on your tasks. The cited lead on five of ten sub-benchmarks does not show whether a specific buyer will save enough to justify the extra cost.
Unestablished return.
The report’s central finding is a twofold list-price difference from Opus 4.8 alongside a 5.7% Intelligence Index lead. Whether that premium yields enough productivity improvement across enterprise work remains unanswered by the public evidence cited.
The Cost of Benchmark Gains
For organizations weighing model costs, the comparison highlights a gap between a clear price premium and evidence that is not yet equally clear about productivity returns. A benchmark lead may matter for a workload that resembles tasks where Fable 5 scores strongly, but an aggregate index cannot establish how much time or money a company will save in its own operations.
The distinction affects procurement decisions. Buyers can calculate token charges from published rates and test model performance on their own tasks. They cannot treat the Stripe account as a general, independently verified estimate of workforce gains. The report’s central finding is not that Fable 5 lacks value, but that available public evidence does not yet show whether the premium produces enough productivity improvement across enterprise work to justify it.
Products Worth Considering
LG UltraGear 27G810A-B 27-Inch 4K UHD IPS Gaming Monitor
Samsung Odyssey G3 G30D 27-Inch FHD Gaming Monitor
LG UltraGear 27GX704A-B 27-Inch QHD OLED Gaming Monitor, 240Hz
As an affiliate, we earn on qualifying purchases.
Pricing and Benchmark Comparisons
According to the source article, Fable 5 launched on June 9, 2026, with general availability around July 1. The report compares it chiefly with Opus 4.8, which it describes as half the price, and Sonnet 4.6, whose listed rates are lower still. It also notes that prompt caching and batch processing have separate discounted rates, so the headline list prices do not represent every customer’s effective cost.
The benchmark figures come from Artificial Analysis, as cited in the report. Its Intelligence Index is an aggregate score, while GDPval-AA is presented as an Elo rating for knowledge-work tasks. The source characterizes the 42-point Elo difference as a narrow percentage when expressed relative to the rating, but says Elo ratings are logarithmic and the raw percentage framing does not directly describe head-to-head performance. These measures answer different questions from a controlled study of how quickly people complete work with or without the model.
“would otherwise have taken a whole team over two months by hand”
— Anthropic, in its launch post as quoted by ThorstenMeyerAI.com
How Broad Are the Productivity Gains?
The report says it found no controlled human-baseline productivity study for Fable 5 and only one public enterprise proof point, the Stripe testimonial. The account does not include publicly reported evaluation methods, independent verification or replication, so readers cannot determine from the cited material how the time comparison was calculated or how broadly it applies.
It is also unclear whether Fable 5’s benchmark advantage translates into better outcomes for a particular company’s work. The report notes stronger results on some sub-benchmarks, but provides no basis for assuming that every organization’s tasks fall into those categories. The listed prices and cited benchmark scores are specific to the source’s July 2026 account; actual costs depend on usage patterns, caching, batch rates and other applicable pricing terms.
Evidence Buyers Can Seek
The next meaningful evidence would include published evaluations that compare work completed with Fable 5 against a defined human or no-AI baseline, describe the methods used, and allow independent review or replication. Until such data appears, organizations considering the model can compare its benchmark performance with their own workloads and measure time, quality and total token costs in a pilot.
Anthropic may publish further customer results or technical evaluations, but the source does not identify a scheduled study or release date. For now, the documented comparison is a twofold list-price difference from Opus 4.8 alongside a 5.7% Intelligence Index lead; the productivity return remains unestablished by the public evidence cited.
Key Questions
How much does Claude Fable 5 cost?
The source reports list prices of $10 per million input tokens and $50 per million output tokens. It also describes lower rates for prompt-cache reads and batch processing, which can change a customer’s effective cost.
How does Fable 5 compare with Opus 4.8?
In the cited comparison, Fable 5 costs twice as much and scores 64.9 on the Intelligence Index, compared with 61.4 for Opus 4.8. The report also lists GDPval-AA scores of 1,932 and 1,890. These are benchmark results, not direct measures of workplace productivity.
Is Fable 5 proven to make enterprise workers more productive?
Not by the public evidence described in the report. It identifies one Anthropic customer testimonial about a Stripe code migration, but says there is no published controlled study, independent audit or replication supporting a general productivity claim.
Could Fable 5 still be worth the higher price?
Possibly, depending on the tasks and results a buyer needs. The report says Fable 5 led on five of ten cited sub-benchmarks, but benchmark performance alone does not show whether the extra cost will pay off for a specific organization. Buyers would need workload-specific testing to judge that.
Source: ThorstenMeyerAI.com
