🔍 Read the full analysis: Paying Twice As Much For 5.7% More AI Intelligence on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
A July 2026 analysis of Anthropic’s Claude Fable 5 finds its listed token prices are twice Opus 4.8’s, while Artificial Analysis scores it 5.7% higher on the Intelligence Index. The analysis says public evidence for enterprise productivity gains consists of one unaudited customer story, not a controlled 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 scoring 5.7% higher on Artificial Analysis’s Intelligence Index, according to a July 2026 review of model pricing and public evidence. The review says the case that Fable 5 raises enterprise productivity is much less established: its public evidence includes one unaudited customer anecdote, not a controlled study against a human-work baseline.
The published prices put Fable 5 at $10 per million input tokens and $50 per million output tokens, compared with $5 and $25 for Opus 4.8. The review places Sonnet 4.6 at about $3 per million input tokens and $15 per million output tokens, making Fable 5 roughly 3.3 times as expensive on those list rates. Anthropic’s pricing documentation and product materials are cited as the primary confirmation of Fable 5’s rates.
On Artificial Analysis’s Intelligence Index, Fable 5 scored 64.9, against 61.4 for Opus 4.8. The review describes that difference as a 5.7% gain. On the separate GDPval-AA knowledge-work Elo measure, the scores were 1,932 and 1,890, respectively. The review cautions that those summary figures do not establish how the models compare on a particular company’s tasks; it says Fable 5 set records in five of the ten sub-benchmarks, with results varying by task.
The only enterprise productivity example identified in the review comes from Anthropic’s launch post. The company attributed to Stripe a claim that Fable 5 completed a migration across a 50-million-line Ruby codebase in one day, a task the account said would otherwise have taken a team more than two months by hand. The review says the statement is evidence that a customer story was published, but it found no disclosed methodology, independent audit or replication to establish the claimed comparison as a measured result.
Paying Twice As Much For 5.7% More AI Intelligence
Claude Fable 5’s list prices are double Opus 4.8’s, while its Artificial Analysis Intelligence Index score is 5.7% higher. Public evidence for enterprise productivity remains one unaudited customer story.
The price ladder
Published per-million-token list rates show a clear premium. Actual spend varies with token mix, caching, batch pricing, and usage.
Claude Fable 5
Claude Opus 4.8
Claude Sonnet 4.6
Capability gains are modest on the index
Artificial Analysis reports a 64.9 score for Fable 5 versus 61.4 for Opus 4.8. This 5.7% difference is a benchmark result, not a productivity estimate.
Intelligence Index
Reported score · higher is better
Estimated full index run
Analysis estimate · benchmark cost, not an enterprise bill
A striking story, not a controlled result
The public productivity claim identified in the review comes from Anthropic’s launch post, which attributes the account to Stripe.
One day versus more than two months
Anthropic said Fable 5 completed a migration across a 50-million-line Ruby codebase in one day, work described as otherwise taking a team over two months by hand.
No disclosed measurement method
The review found no disclosed study design, human comparison group, quality checks, independent audit, or replication to verify the claim as a measured result.
“Fable 5 completed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.” Anthropic launch post, citing Stripe
How the illustrative rate adds up
Fable 5 offers lower prices under specific cache and batch conditions. The blended figure below reflects one stated usage mix, not a universal rate.
A cache-heavy example
Prompt-cache reads are listed at $1 per million tokens, a 90% discount. Batch rates are $5 input and $25 output per million tokens. US-only inference carries a 1.1× multiplier, or $11 and $55.
7:2:1 cached reads : input : output
A better buyer test
Compare models on the same real tasks, then track the full cost and quality of the work.
Match tasks
Use representative company workloads.
Measure output
Record completion time and quality.
Count review
Include human checking and rework.
Compare total
Add token costs and integration effort.
Key questions
The review describes available public evidence as limited. Its conclusion is an evidence gap, not proof that Fable 5 cannot improve productivity.
What does Fable 5 cost?
$10 per million input tokens and $50 per million output tokens at listed rates. Cache and batch pricing can reduce charges when eligible.
How does it compare with Opus 4.8?
Fable 5 costs twice as much at list rates. Its Intelligence Index score is 64.9 versus 61.4.
Does 5.7% more index score mean 5.7% more productivity?
No. The percentage describes the benchmark score difference; it does not measure workplace productivity.
What evidence would help?
Workload-specific comparisons with documented baselines, time, quality, review effort, and total token costs.
The pricing and benchmark comparison matter to organizations deciding whether to pay more for a top-end model. At list rates, a Fable 5 workload using the same number and mix of tokens as an Opus 4.8 workload would carry twice the model cost. Whether the higher score justifies that premium depends on the work being done, not on the aggregate index alone.
The review also estimates that a full Intelligence Index run costs about $9,940 for Fable 5 versus $4,970 for Opus 4.8. That is the estimated cost of the benchmark run, not a forecast of an enterprise’s operating bill or a measure of return on investment. In practice, prompt caching, batch pricing and the balance of input and output tokens can affect costs. The source calculates a blended rate of $7.70 per million tokens for Fable 5 under a stated cache-hit, input and output mix; actual usage may differ.
For buyers, the central distinction is between model capability scores and demonstrated workplace productivity. Benchmarks can help compare performance on defined tests. They do not, by themselves, show that a model will complete a company’s work faster, reduce staffing needs or save money after integration and review costs. The available customer story may point to a promising use case, but the review says it is not enough to quantify those outcomes broadly.
Popular right now
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Cypress Green
Bose QuietComfort Wireless Bluetooth Headphones – Active Over‑Ear Noise Cancelling with Mic, USB‑C Charging, Deep Bass, up to 24 h Playtime,
TCL 65-Inch 4K UHD Smart LED TV (65S425)
As an affiliate, we earn on qualifying purchases.
How the Price Comparison Was Built
The review says Fable 5’s list prices are corroborated by Anthropic’s launch post, platform pricing documentation and product page, as well as secondary reporting and pricing sources. It also notes that Fable 5 offers a 90% prompt-caching discount, with cache reads priced at $1 per million tokens, and batch pricing at $5 per million input tokens and $25 per million output tokens. US-only inference is listed with a 1.1-times price multiplier, or $11 and $55 per million tokens.
Its blended estimate uses a 7:2:1 mix of cached reads, ordinary input and output tokens: (7 × $1 + 2 × $10 + 1 × $50) divided by 10 equals $7.70 per million tokens. That is an illustrative calculation based on the specified mix, rather than a universal rate. The review also says Fable 5 and Mythos 5 share the same underlying weights and list prices, while differing in safeguards.
The comparison is with Opus 4.8, priced at half Fable 5’s list rates, and with Sonnet 4.6 as a less expensive active-lineup option. The review calls a separate comparison with GPT-5.5 directional because it relies on a single secondary source. It reports Fable 5’s launch on June 9, 2026, and general availability around July 1.
“Fable 5 completed a codebase-wide migration in a day that “would otherwise have taken a whole team over two months by hand.””
— Anthropic’s launch post, citing Stripe
Productivity Claims Lack Independent Testing
The public material reviewed does not establish whether Fable 5 produces a measurable productivity gain across enterprise knowledge work. The analysis reports no controlled human-baseline productivity study for the model and one unaudited customer example. It does not provide the study design, task selection, human comparison group, quality checks or independent replication needed to assess the Stripe account as a general measure of productivity.
It is also unclear how the index and sub-benchmark differences translate to any particular customer’s workloads. The source describes stronger results on five of ten sub-benchmarks, but does not establish that a given organization’s tasks match those tests. The review’s estimated costs depend on stated assumptions about benchmark usage or token mix; they are not actual customer bills.
The review is an analysis of available public evidence, not a controlled evaluation of Fable 5 inside a company. Its conclusion about the evidence gap should not be read as proof that the model cannot improve productivity. It means the public information cited does not yet quantify or independently verify such a gain.
Test Costs Against Actual Work
For organizations weighing the premium, the next useful evidence would be workload-specific testing that compares Fable 5 with less expensive models on the same tasks. A decision would be better grounded if tests reported completion time, output quality, human review effort and total token costs, and used a documented baseline. The source does not identify a forthcoming independent productivity study or a scheduled next benchmark release.
Anthropic’s listed prices and product availability provide a basis for cost estimates, but buyers will need their own usage mix to estimate bills. The published cache and batch rates can lower token charges under eligible conditions; they do not establish whether the higher-priced model will save money overall. Until more independently documented evidence appears, the productivity case remains a customer anecdote rather than a verified general result.
Key Questions
How much does Claude Fable 5 cost?
Anthropic lists Fable 5 at $10 per million input tokens and $50 per million output tokens. The source also reports lower rates for prompt-cache reads and batch processing, subject to those pricing conditions.
How does it compare with Opus 4.8?
Fable 5’s listed input and output rates are twice Opus 4.8’s, which the source gives as $5 and $25 per million tokens. Artificial Analysis scores Fable 5 at 64.9 and Opus 4.8 at 61.4 on its Intelligence Index.
Does the 5.7% benchmark gain prove Fable 5 boosts productivity by 5.7%?
No. The 5.7% figure refers to the difference on the Intelligence Index, not a measured increase in workplace productivity. The review says it found no controlled human-baseline productivity study for Fable 5.
What is the evidence behind the Stripe productivity claim?
Anthropic’s launch post attributes to Stripe a claim that Fable 5 migrated a 50-million-line Ruby codebase in one day, work described as taking a team more than two months by hand. The review found no public audit, methodology or independent replication for that account.
Could Fable 5 still be worth the higher price?
Possibly for tasks where its performance advantage is useful, but the source does not establish a general return on investment. Buyers would need to compare models on their own tasks, quality requirements and usage costs.
Source: ThorstenMeyerAI.com
