TL;DR
Anthropic’s Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, twice the cost of Opus 4.8. Third-party benchmark data cited in the source material shows a 5.7% Intelligence Index gain, while public productivity evidence rests on one unaudited customer story.
Anthropic’s Claude Fable 5 is available at $10 per million input tokens and $50 per million output tokens, double the price of Opus 4.8, while independent benchmark data cited in the source material shows only a 5.7% higher Intelligence Index score. The gap matters for enterprise buyers because the model’s publicly documented productivity case remains thin.
Confirmed pricing is the clearest part of the story. Anthropic lists Fable 5 at $10 input and $50 output per million tokens, compared with Opus 4.8 at $5 and $25. The source material says that pricing is also reflected in Anthropic’s product and platform materials and has been repeated by outlets including Forbes, TechCrunch, Vellum, Finout and Artificial Analysis.
Artificial Analysis benchmarked Fable 5 at 64.9 on its Intelligence Index, compared with 61.4 for Opus 4.8. That is the basis for the 5.7% measured gain cited in the source material. On GDPval-AA, a knowledge-work benchmark, Fable 5 is listed at 1,932 Elo versus 1,890 for Opus 4.8, a smaller aggregate spread in raw percentage terms.
The productivity claim is less settled. Anthropic’s launch post credits Stripe with using Fable 5 on a 50-million-line Ruby codebase migration in one day, saying the work would otherwise have taken a team more than two months by hand. That is confirmed as a vendor-published customer claim, not as an audited enterprise study.
Budget Pressure From Token Costs
For AI buyers, the issue is not whether Fable 5 is stronger than Opus 4.8; the cited public benchmark says it is. The harder question is whether a 2x price premium pays for itself when the aggregate intelligence gain is 5.7%.
The answer may vary by workload. The source material says Fable 5 set records in five of 10 sub-benchmarks, and a 42-point Elo gap can imply a higher head-to-head win rate than a simple percentage suggests. Even so, the public evidence does not yet show broad, measured productivity gains across enterprise teams.
AI Translation Earbuds Real Time,199 Language Translator Earbuds,Bluetooth 6.1 Audifonos Traductores Inglés Español,4 ANC Mic Translating Device,Translate Ear Buds for Travel Learning with LCD Touch
🎵𝗥𝗲𝗮𝗹-𝗧𝗶𝗺𝗲 𝗔𝗜 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗶𝗼𝗻 𝗘𝗮𝗿𝗯𝘂𝗱𝘀: 𝗕𝗿𝗲𝗮𝗸 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗕𝗮𝗿𝗿𝗶𝗲𝗿𝘀 𝗶𝗻 𝟭𝟵𝟵 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀 𝗜𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆 — Experience seamless global communication with these…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Fable 5 Is Priced
Fable 5 launched on June 9, 2026, with general availability described in the source material as arriving around July 1, 2026. Its price places it above Anthropic’s active lineup: Opus 4.8 is half the cost, while Sonnet 4.6 is cited at roughly one-third of the price.
The effective bill can be lower than the sticker price. The source material lists a 90% prompt-caching discount, a batch tier at $5 input and $25 output per million tokens, and a blended example of $7.70 per million tokens using a 7:2:1 cache-hit-to-input-to-output mix. It also says Fable 5 and Mythos 5 use the same underlying weights, with safeguards as the key difference.
“would otherwise have taken a whole team over two months by hand”
— Anthropic launch post, credited to Stripe
The Self-Taught Cloud Computing Engineer: A comprehensive professional study guide to AWS, Azure, and GCP
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Productivity Proof Is Still Thin
It is not yet clear whether the Stripe example reflects repeatable gains for other companies, codebases or teams. The source material says there are zero controlled human-baseline productivity studies and only one unaudited customer proof point in the public record.
Several cost comparisons also depend on deployment details. Real spending will vary with cache-hit rates, output length, task mix, latency needs and quality thresholds. The source material treats the GPT-5.5 price comparison as directional because it rests on a single secondary source.
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Enterprise Tests Become The Gate
The next milestone is evidence from real enterprise workloads, not only leaderboard averages. Buyers are likely to compare Fable 5 against Opus 4.8 and cheaper models on their own tasks, then measure whether higher accuracy or speed offsets the token-price premium.
The strongest new evidence would be audited productivity studies, public workload-specific evaluations, or detailed customer reports with baseline methods. Until that appears, claims about broad productivity gains remain vendor-attributed and unproven at scale.
Performance Evaluation and Benchmarking: 12th TPC Technology Conference, TPCTC 2020, Tokyo, Japan, August 31, 2020, Revised Selected Papers (Programming and Software Engineering)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news about Claude Fable 5?
Claude Fable 5 is now available at $10 per million input tokens and $50 per million output tokens. That is twice the listed price of Claude Opus 4.8.
Is Fable 5 better than Opus 4.8?
On the cited Artificial Analysis benchmark, yes. Fable 5 scored 64.9 versus 61.4 for Opus 4.8, a 5.7% aggregate gain.
Does the Stripe example prove major productivity gains?
No. The Stripe migration is a public claim from Anthropic’s launch material. The source material says it is unaudited and not backed by controlled human-baseline studies.
Can prompt caching reduce the cost?
Yes. The source material lists a 90% prompt-caching discount and a blended example of $7.70 per million tokens. Actual savings depend on how often prompts repeat and how much output the model generates.
What should buyers watch next?
Watch for independent workload tests, audited customer studies and direct comparisons with Opus 4.8 on production tasks. Those results will matter more for budgets than a single aggregate leaderboard score.
Source: Thorsten Meyer AI
