🔍 Read the full analysis: A Small AI Intelligence Gain Comes At Double 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 is listed at $10 per million input tokens and $50 per million output tokens, twice the published rates for Opus 4.8. A July analysis by ThorstenMeyerAI.com says third-party benchmark data show a 5.7% lead over Opus 4.8 on Artificial Analysis’s Intelligence Index, while the public case for greater workplace productivity rests on one unaudited customer account rather than a controlled study.

The comparison sets Fable 5’s Intelligence Index score at 64.9, against 61.4 for Opus 4.8. On GDPval-AA, a benchmark the analysis describes as measuring knowledge-work performance, Fable 5 scored 1,932 Elo to Opus 4.8’s 1,890, a 42-point difference. The report cautions that the two benchmark results describe different measures and should not be treated as direct evidence of productivity gains at a company.

Artificial Analysis’s full Intelligence Index run is estimated in the source at about $9,940 for Fable 5 and $4,970 for Opus 4.8. Those are benchmark-run costs, not a prediction of what every customer will pay: actual bills depend on usage and pricing arrangements. The report also says Fable 5 records the highest result in five of the index’s 10 sub-benchmarks, suggesting performance varies by task.

The productivity example cited in Anthropic’s launch post concerns a 50-million-line Ruby codebase. Anthropic attributed the account to Stripe, saying Fable 5 completed a codebase-wide migration in a day that “would otherwise have taken a whole team over two months by hand.” The analysis found no published methodology, independent audit or replication for that account. It treats the statement as a customer testimonial, not as a measured productivity result.

The Premium Depends on the Workload

For organizations choosing a model, the figures frame a practical procurement question: whether Fable 5’s performance on a particular task can justify its two-times list price relative to Opus 4.8. A broad benchmark average may not predict results on a company’s own work. The analysis says Fable 5 leads on five of 10 sub-benchmarks, so teams whose tasks match those areas could see a different value calculation from teams whose work does not.

The gap between benchmark scores and workplace outcomes matters because model spending is often justified in terms of staff time saved or work completed faster. The source identifies no controlled human-baseline productivity study for Fable 5. That does not establish that the model lacks productivity benefits; it means the cited public material does not quantify them in a controlled, independently checked way. Buyers relying on the Stripe account alone cannot know whether the same result applies to other firms, codebases or kinds of work.

Pricing is more straightforward to compare, though list rates are not the whole bill. The analysis cites prompt caching, batch pricing and a usage-based blended-rate estimate that can lower costs in some circumstances. Customers therefore need to compare both their expected workload and the applicable pricing tier, rather than assume either the headline rate or benchmark average represents their own costs and returns.

How the Price and Scores Compare

According to the analysis, Anthropic’s published rates place Fable 5 at $10 per million input tokens and $50 per million output tokens. Opus 4.8 is listed at $5 and $25, respectively; Sonnet 4.6 is listed at roughly $3 and $15. Fable 5 is therefore the most expensive model in Anthropic’s active lineup described by the source. The article says the price is corroborated by Anthropic materials and several secondary outlets.

The headline rate can differ from effective costs. The source lists cache reads at $1 per million tokens, a 90% prompt-caching discount, and batch rates of $5 per million input tokens and $25 per million output tokens. It estimates a blended rate of $7.70 per million tokens under a specific 7:2:1 ratio of cache-hit, input and output tokens. That is an illustrative calculation, not a universal customer rate. US-only inference is listed at a 1.1-times multiplier.

The report also notes that Fable 5 and Mythos 5 share the same underlying weights and are priced identically, with safeguards differing. Its comparison focuses on Fable 5 versus Opus 4.8 because Opus 4.8 is presented as half the price. The 5.7% figure refers specifically to the Intelligence Index score difference; it is not a claim that Fable 5 makes workers 5.7% more productive.

“would otherwise have taken a whole team over two months by hand”

— Anthropic’s launch post, quoting a Stripe customer account as reported by ThorstenMeyerAI.com

Productivity Claims Lack Independent Testing

The source analysis reports zero controlled human-baseline productivity studies for Fable 5 and one enterprise proof point: the Stripe account included in Anthropic’s launch post. The available material does not explain how the time estimate was calculated, what human work was included, what review or correction the model’s output required, or whether the result was independently verified.

It is also unclear how well the benchmark differences predict outcomes in specific organizations. The source gives aggregate scores and sub-benchmark placements but does not provide evidence that every customer workload will show the same gains. The analysis describes its benchmark figures as third-party data; readers should distinguish those measured scores from claims about business impact.

The source gives an expected general-availability date of around July 1, 2026, following a June 9 announcement. It does not establish in the supplied material whether that rollout occurred as scheduled. It also does not provide customer-specific discounts, typical token usage or enough information to calculate a general return on investment.

Watch for Audited Customer Results

The next useful evidence would be independently reviewed productivity studies that compare Fable 5 with a defined human or no-AI baseline. Such work would need to explain the task, time period, review requirements and costs, rather than relying only on a customer estimate. The source material does not announce a date for such research or identify a planned audit.

In the meantime, buyers can compare the model’s results on tasks close to their own work and calculate costs using their expected input, output and cached-token mix. The cited benchmark record suggests the premium may be more defensible for some tasks than others, but the supplied evidence does not settle that question for individual organizations.

Key Questions

How much does Claude Fable 5 cost?

The source lists $10 per million input tokens and $50 per million output tokens. It also describes lower batch rates and a prompt-caching discount; a customer’s bill depends on usage and applicable pricing.

How much higher is Fable 5’s Intelligence Index score?

Artificial Analysis scores Fable 5 at 64.9 and Opus 4.8 at 61.4, a 5.7% difference as calculated in the source analysis. This benchmark comparison is not a measurement of worker productivity.

Is there evidence that Fable 5 improves workplace productivity?

The source identifies one customer account, attributed to Stripe and included in Anthropic’s launch post. It reports no controlled human-baseline study, independent audit or replication for the claim.

Does the benchmark show Fable 5 is better for every task?

No. The analysis says Fable 5 ranked highest in five of the Intelligence Index’s 10 sub-benchmarks. It does not establish that every task or customer will see the same advantage.

Source: ThorstenMeyerAI.com