TL;DR
Anthropic’s Claude Fable 5 is the company’s most expensive active public model at $10 per million input tokens and $50 per million output tokens. Third-party benchmark data cited in the source material shows it scoring 5.7% higher than Opus 4.8 on the Artificial Analysis Intelligence Index, while public productivity evidence rests on one unaudited customer account.
Anthropic’s Claude Fable 5 has entered public availability as the company’s highest-priced active Claude model, charging $10 per million input tokens and $50 per million output tokens, while benchmark data cited by Artificial Analysis shows a 5.7% Intelligence Index gain over the half-priced Opus 4.8.
The pricing is the most firmly documented part of the story. According to the source material, Anthropic’s launch post, platform pricing documents and product page list Fable 5 at $10/$50 per million input and output tokens. That is 2x Opus 4.8, listed at $5/$25, and about 3.3x Sonnet 4.6, listed at roughly $3/$15.
The performance comparison is narrower. Artificial Analysis benchmark data cited in the material gives Fable 5 an Intelligence Index score of 64.9, compared with 61.4 for Opus 4.8, a 5.7% increase. On GDPval-AA, a knowledge-work benchmark, Fable 5 scored 1,932 Elo against 1,890 for Opus 4.8, a 42-point gap.
The source material also says the economics soften under some usage patterns. Fable 5 includes a 90% prompt-caching discount, cache reads priced at $1 per million tokens, a $5/$25 batch tier, and a blended rate of $7.70 per million tokens under a 7:2:1 cache-hit, input and output mix. US-only inference is listed at a 1.1x multiplier.
Budget Pressure for AI Buyers
The gap matters because enterprise AI budgets are increasingly tied to measurable return. A buyer paying 2x the Opus 4.8 price is not only buying a higher benchmark score; they are betting that Fable 5’s task-specific gains will offset a much higher unit cost in production workloads.
The evidence is mixed rather than one-sided. The source material says Fable 5 led five of 10 sub-benchmarks, meaning some difficult tasks may show larger gains than the aggregate score suggests. At the same time, the public data cited does not show a broad 2x performance improvement matching the 2x price premium.
Unlocking Data with Generative AI and RAG: Enhance generative AI systems by integrating internal data with large language models using RAG
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Fable Compares With Opus
Fable 5 now sits at the top of Anthropic’s active public lineup by price, according to the material. Older Opus 4 and Opus 4.1 models were listed at $15/$75, but the source describes them as retired, leaving Fable 5 as the costliest active option.
The source also clarifies that Fable 5 and Mythos 5 use the same underlying weights and carry the same pricing, differing in safeguards. That means price comparisons that group them together are treated in the material as accurate rather than a category error.
A comparison with GPT-5.5 is less firm. The material says Fable 5’s input price is roughly 2x GPT-5.5’s, but also says that comparison relies on one secondary source and should be read as directional rather than fully established.
“would otherwise have taken a whole team over two months by hand”
— Anthropic, citing Stripe
Subscription Business Construction Series Vol 1 Business Model Design: Design Your Subscription Revenue Model Target and Pricing from Zero (Hironext Publishing) (Japanese Edition)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Productivity Claims Still Thin
The largest open question is whether Fable 5 produces enough real-world productivity to justify its price. The source material says it found zero controlled human-baseline productivity studies and only one enterprise proof point, the Stripe migration account cited by Anthropic.
That Stripe example survived only as a claim that was made, not as independently verified measurement. The material says stronger descriptions of the case as concrete productivity proof or an independently established baseline did not survive verification.
It is also unclear which workloads benefit most from Fable 5. The benchmark spread suggests task-level variation, so the model may be more compelling for some hard tasks than for broad everyday use.
Evals for AI Engineers: Systematically Measuring and Improving AI Applications
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Procurement Tests Move Center Stage
The next step for buyers is likely internal evaluation rather than reliance on launch claims. Teams weighing Fable 5 will need to test their own prompts, cache-hit rates, output volume and human review time against Opus 4.8 and cheaper Claude models.
More public evidence may come from customer case studies, independent audits or benchmark updates. Until then, the confirmed story is a premium-priced model with top-line benchmark gains, while the broader productivity case remains still developing.
How LLMs Actually Work: Large Language Models Explained: From Hardware to Hallucination, Tokens to Transformers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much does Claude Fable 5 cost?
Claude Fable 5 is listed at $10 per million input tokens and $50 per million output tokens. The source material says that is double the price of Opus 4.8.
Is Fable 5 better than Opus 4.8?
On the cited Artificial Analysis Intelligence Index, Fable 5 scored 64.9 compared with 61.4 for Opus 4.8. That is a 5.7% aggregate gain, though some sub-benchmarks showed larger task-specific differences.
Is there proof Fable 5 makes workers more productive?
The public evidence cited here is limited. The material identifies one unaudited customer story from Stripe, cited by Anthropic, and no controlled human-baseline productivity studies.
Why does the 5.7% figure matter?
The figure matters because Fable 5 costs 2x Opus 4.8 while showing a much smaller aggregate benchmark gain. Buyers may still find value if their workload matches the tasks where Fable 5 performs best.
What should companies check before buying Fable 5?
Companies should compare real workload results, token costs, prompt-caching savings and human review time against cheaper alternatives before committing production budgets.
Source: Thorsten Meyer AI
