TL;DR
Anthropic’s Claude Fable 5 is now positioned as its most expensive active public model, priced at $10 per million input tokens and $50 per million output tokens. Artificial Analysis benchmark data cited in the source material shows a 5.7% Intelligence Index gain over Opus 4.8, which costs half as much, while public productivity evidence rests on one unaudited customer story.
Anthropic’s Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, exactly double Opus 4.8, while third-party benchmark data cited by the source material shows only a 5.7% Intelligence Index gain over that lower-priced model.
Anthropic’s pricing materials, as described in the source material, list Fable 5 at $10/$50 per million tokens. Opus 4.8 is listed at $5/$25, while Sonnet 4.6 was about $3/$15, making Fable 5 the most expensive model in Anthropic’s active public lineup.
The effective rate can be lower for some users. The source material says Fable 5 includes a 90% prompt-caching discount, batch pricing at $5/$25, and a modeled blended rate of $7.70 per million tokens using a 7:2:1 cache-hit, input, and output mix. US-only inference is listed at a 1.1x multiplier.
On performance, Artificial Analysis data cited in the source material gives Fable 5 an Intelligence Index score of 64.9, compared with 61.4 for Opus 4.8. On GDPval-AA, a knowledge-work Elo benchmark, Fable 5 scored 1,932 against 1,890 for Opus 4.8. Those are historical benchmark results, not a guarantee of business return.
Budget Pressure From Fable 5
The gap matters because many AI buyers pay higher model prices only when the gains show up in production workloads. The public record described in the source material shows documented pricing and measured benchmark gains, but much thinner proof for enterprise productivity.
That does not mean Fable 5 has no case. The same source material says it set records in 5 of 10 sub-benchmarks, and a 42-point Elo lead can imply a stronger head-to-head result than a simple 2.2% percentage reading suggests. The issue for buyers is narrower: whether their own use case falls into the tasks where Fable 5’s premium shows up.
MLOps for Decision Makers: The First Plain-Language Guide to Machine Learning Operations — Why 95% of AI Projects Fail and How to Make Sure Yours Doesn't (The AI Literacy Series Book 4)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From June Launch To GA
Fable 5 launched on June 9, 2026, with general availability reported around July 1, 2026. The source material says Fable 5 and Mythos 5 use the same underlying weights and have identical pricing, with differences tied to safeguards rather than model cost.
Anthropic has priced some past Claude models higher. The source material lists retired Opus 4 and 4.1 at $15/$75 per million tokens. Within the active lineup described, though, Fable 5 sits at the top of the public price range.
AI benchmark performance tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Missing Productivity Baseline
It is not yet clear whether the Stripe example was independently audited, how the human baseline was measured, how many engineers were counted, or how review and defect rates were handled. The source material says it found zero controlled human-baseline productivity studies and one unaudited enterprise proof point.
It is also unclear how much private discounts, enterprise contracts, caching patterns, and workload mix change the real cost for large buyers. The listed prices are confirmed in the source material, but actual customer economics may differ from public sticker pricing.
AI model input output tokens
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Pilot Results Will Set Value
The next test is whether buyers, benchmarkers, or Anthropic publish audited productivity data tied to real knowledge-work baselines. Until then, Fable 5’s case rests on higher benchmark scores, top-line pricing, and limited public evidence about how much work it saves.
For teams setting 2026 AI budgets, the practical next step is workload-level testing against Opus 4.8, cheaper Claude models, and rival systems. This article is not financial, tax, or legal advice; model prices and benchmark results are historical inputs for evaluation, not guarantees of future return.
AI model prompt caching tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the confirmed price of Claude Fable 5?
The source material says Claude Fable 5 is listed at $10 per million input tokens and $50 per million output tokens, with lower effective rates possible through caching and batch pricing.
How much better is Fable 5 than Opus 4.8 in the cited benchmark?
Artificial Analysis data cited in the source material shows Fable 5 at 64.9 on the Intelligence Index versus 61.4 for Opus 4.8, a 5.7% gain.
Does the 5.7% gain mean Fable 5 is never worth the price?
No. The source material says Fable 5 led in 5 of 10 sub-benchmarks, so some workloads may benefit more than the blended score suggests. The value depends on specific tasks, token mix, caching, and error costs.
What public productivity proof exists for Fable 5?
The source material identifies one surviving enterprise proof point: an Anthropic-cited Stripe migration claim. It also says it found no controlled human-baseline productivity studies in the public evidence base.
Source: Thorsten Meyer AI
