🔍 Read the full analysis: A 5.7% Edge In AI Intelligence Comes At Twice The Price on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

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 pricing and benchmark figures reviewed by ThorstenMeyerAI.com. The performance gain is measured on benchmark tests; the public evidence that Fable 5 increases enterprise productivity is much thinner, consisting of one unaudited customer account rather than a controlled study.

Anthropic lists Fable 5 at $10 per million input tokens and $50 per million output tokens. Opus 4.8 is listed at $5 and $25, respectively, while Sonnet 4.6 is listed at about $3 and $15. That makes Fable 5 the most expensive model in Anthropic’s active lineup, based on the supplied review. The review says the pricing is corroborated by Anthropic’s own product and platform materials.

Artificial Analysis scores Fable 5 at 64.9 on its Intelligence Index, compared with 61.4 for Opus 4.8, a 5.7% difference as reported by the source. On the separate GDPval-AA knowledge-work benchmark, Fable 5 scores 1,932 Elo against Opus 4.8’s 1,890, a reported 2.2% gap. A full Intelligence Index run is estimated to cost about $9,940 for Fable 5 and $4,970 for Opus 4.8; those figures are benchmark-run costs, not typical customer bills.

The source review identifies one public enterprise productivity example: Anthropic’s launch post quoted Stripe as saying Fable 5 completed a migration across a 50-million-line Ruby codebase in a day, a task that would otherwise have taken a team more than two months by hand. The review says this survives verification as a customer claim that was published, but not as an independently audited result. It found no controlled human-baseline productivity study for Fable 5 in the public material it examined.

The Premium Depends on the Workload

The comparison matters to organizations choosing models for large-scale use because token pricing can compound across repeated tasks, while benchmark scores do not directly predict the time or money saved in a particular workplace. At list price, Fable 5’s higher score comes with a twofold token-price premium over Opus 4.8. Whether that trade is worthwhile depends on how often the more capable model succeeds where a cheaper one does not, and on the cost of errors, review, and reruns.

The benchmark results also do not show that the advantage is evenly distributed. ThorstenMeyerAI.com reports that Fable 5 set records in five of ten sub-benchmarks, with larger gains on some tasks than the overall index suggests. That makes the aggregate 5.7% figure an incomplete guide for buyers whose work is concentrated in a particular category. The supplied material does not provide enough task-level detail to determine which customers would see a return.

Pricing is more directly verifiable than the productivity case. The source gives a blended estimate of $7.70 per million tokens under a specific mix: seven cache-read tokens, two ordinary input tokens and one output token. That is an illustrative calculation using a 90% prompt-cache discount, not a universal effective rate. Actual costs will vary with the balance of cached input, uncached input and output.

From Launch to Public Evidence

The source dates Fable 5’s launch to June 9, 2026, with general availability around July 1. Its review focuses on a narrow purchasing question: what is established about the model’s price and benchmark performance, and what has been demonstrated about productivity compared with human work? It distinguishes Anthropic’s published price information from claims about customer outcomes.

The listed prices are not the only available rates. The review says prompt-cache reads cost $1 per million tokens, batch processing is priced at $5 per million input tokens and $25 per million output tokens, and US-only inference carries a 1.1× multiplier, or $11 and $55. It also says Fable 5 and Mythos 5 use the same underlying weights and have identical prices, with safeguards differing between them. These options can change the bill, but do not establish how much labor a customer saves.

For the benchmark comparison, the review relies on Artificial Analysis results and reports both the Intelligence Index and GDPval-AA scores. Those are distinct measures: the index is an aggregate benchmark score, while GDPval-AA reports an Elo rating for knowledge-work tasks. The source cautions against treating a percentage conversion of the Elo gap as a complete account of practical performance; it says a 42-point lead corresponds to roughly a 56% head-to-head win rate under the rating interpretation it cites.

“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

Productivity Gains Remain Unmeasured

The available material does not establish that Fable 5 makes workers or organizations more productive by a particular percentage. The Stripe example lacks a published methodology, independent audit and replication, according to ThorstenMeyerAI.com’s review. It supports reporting that Anthropic cited a customer account, but does not establish a general productivity result.

It is also unclear how closely benchmark performance maps to the tasks, software, oversight requirements and costs of any given customer. The supplied review gives aggregate scores and says five of ten sub-benchmarks were records, but does not provide enough detail here to identify the relevant advantage for a specific deployment. The reported blended price likewise depends on an assumed token mix and should not be read as a guaranteed rate.

The source review reports that it checked 25 claims, but the supplied material does not include the underlying audit records or full benchmark methodology. Readers should treat its descriptions of those checks as the review’s account, while attributing the benchmark scores to Artificial Analysis and the customer story to Anthropic’s launch post.

Buyers Need Task-Level Tests

For organizations considering Fable 5, the immediate next step is to compare it with a lower-cost model on representative tasks rather than infer a business case from an aggregate score. A useful evaluation would track task completion, human review time, error rates and total token costs against a defined baseline. Those results would show whether the higher price is justified for that workload; the supplied source does not report such a study.

Further public evidence could change the assessment if Anthropic or customers publish audited, repeatable productivity measurements with clear methods and comparison groups. Until then, the available record supports a precise price comparison and benchmark advantage, but not a broad claim about labor savings. No additional study or publication date is specified in the source material.

Key Questions

How much does Claude Fable 5 cost?

The supplied July 2026 source lists Fable 5 at $10 per million input tokens and $50 per million output tokens. Prompt-cache reads and batch processing have different listed rates.

What does the 5.7% figure measure?

It is the reported difference between Fable 5’s 64.9 Intelligence Index score and Opus 4.8’s 61.4 score on Artificial Analysis. It is a benchmark comparison, not a measured productivity increase.

Is Fable 5 twice as productive as Opus 4.8?

No such result is established by the supplied material. Fable 5 costs twice as much per token, but the source review found no controlled human-baseline productivity study for the model.

What is the Stripe example?

Anthropic’s launch post quoted Stripe as saying Fable 5 completed a migration across a 50-million-line Ruby codebase in one day, a task that would otherwise have taken a team more than two months. The source review describes this as an unaudited customer testimonial, not an independently verified measurement.

Could the higher price still make sense for some buyers?

Possibly, if a buyer’s tasks are among those where Fable 5 performs better and the improvement reduces other costs. The supplied review says the model set records in five of ten sub-benchmarks, but task-level results and customer-specific cost savings are not established here.

Source: ThorstenMeyerAI.com