TL;DR
Anthropic’s Claude Fable 5 is priced at $10/$50 per million input/output tokens, double Opus 4.8. Artificial Analysis benchmark data cited in the source puts its Intelligence Index at 64.9 versus 61.4 for Opus 4.8, a 5.7% gain, while public productivity evidence rests on one unaudited Stripe story carried by Anthropic.
Anthropic’s Claude Fable 5, launched on June 9, 2026 and reported generally available around July 1, 2026, costs $10 per million input tokens and $50 per million output tokens, while a third-party benchmark cited in the source shows a 5.7% Intelligence Index gain over Opus 4.8 at twice the list price, a cost question for buyers setting 2026 AI budgets.
Fable 5’s list price is documented at $10 per million input tokens and $50 per million output tokens in Anthropic’s launch materials, pricing docs and product page, according to the source material. That is 2.0 times Opus 4.8 at $5/$25 and about 3.3 times Sonnet 4.6 at roughly $3/$15, with Forbes, TechCrunch, Vellum, Finout and Artificial Analysis cited as corroborating outlets.
The headline rate is softened by discounts that Anthropic offers for some usage patterns. The source cites a 90% prompt-caching discount, cache reads at $1 per million tokens, a $5/$25 batch tier and an effective blended rate of $7.70 per million tokens under a 7:2:1 cache-hit, input and output mix.
Artificial Analysis put Fable 5 at 64.9 on its Intelligence Index, compared with 61.4 for Opus 4.8, a 5.7% increase. On GDPval-AA, the source reports 1,932 Elo for Fable 5 versus 1,890 for Opus 4.8; the full Index run is estimated at about $9,940 for Fable 5 and $4,970 for Opus 4.8.
The budget issue is straightforward: buyers are being asked to pay a 2x list-price premium for an aggregate benchmark gain that is real but limited in the data cited. For teams buying at scale, the difference can move from a model-choice debate to a large operating-cost decision.
The performance picture is less flat than the 5.7% headline suggests. The source says Fable 5 led in five of 10 sub-benchmarks, and a 42-point Elo gap on GDPval-AA can translate into a stronger task-level advantage than a plain percentage implies. That means the value case depends on whether a buyer’s workload falls in the areas where Fable 5’s gains appear strongest.
The Hidden Cost of AI: A Practical Handbook for Token Optimisation: A Practical Handbook for Token and Credit Optimisation in Production LLM Systems (AI Engineering Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Fable 5 Compares
Fable 5 sits above Opus 4.8 in the active Claude lineup on both price and benchmark score. The source says Opus 4 and 4.1 were the only Claude models priced higher at $15/$75 per million tokens, while Sonnet 4.6 is listed as retired in the comparison.
Fable 5 and Mythos 5 are described as the same underlying weights with different safeguards, so reports that group them together on price are not conflating separate products. The source also cites US-only inference at a 1.1x multiplier, or $11/$55 per million tokens, for customers with data-residency needs.
“In a 50-million-line Ruby codebase, Fable 5 performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.”
— Anthropic launch post, citing Stripe
AMT Star Trek Model Kit (U.S.S. Enterprise Refit)
Plastic Model Kit, Assembly required
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Productivity Evidence Still Thin
The main open issue is not whether Fable 5 scores higher in published benchmarks; the source says it does. The open issue is whether it delivers enterprise productivity gains large enough to justify the premium, because the public record cited here includes no controlled human-baseline studies.
The surviving human-baseline proof point is a Stripe example carried in Anthropic’s launch post. Anthropic said Fable 5 completed a migration in a 50-million-line Ruby codebase in one day that would have taken a whole team over two months by hand, but the source describes that as unaudited customer evidence, not an independently verified study.
NextGen Tech Breakdown – AI Tools, GPU Benchmarks & Web Performance Experiments
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Benchmarks Move To Workloads
Buyers evaluating Fable 5 will likely focus next on private pilots, workload-specific tests and invoice data rather than launch claims. The clearest near-term signal will be whether independent users can show repeatable task gains, measured labor savings and cost per completed job that outperform Opus 4.8 or lower-priced models.
For Anthropic, the next milestone is evidence. More public benchmarks, customer audits or case studies with clear baselines would help show whether Fable 5’s premium is a narrow fit for hard workloads or a broader upgrade for knowledge-work deployments.
Hybrid AI Systems with Small and Large Language Models: Design Cost-Efficient, Scalable, and Secure AI Applications for Edge and Cloud
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the actual news development?
Claude Fable 5 has moved from launch to reported general availability while carrying Anthropic’s highest active Claude list price. The news focus is the gap between double Opus 4.8’s price and a 5.7% aggregate benchmark gain.
How much does Claude Fable 5 cost?
Anthropic lists Fable 5 at $10 per million input tokens and $50 per million output tokens. The source also cites batch and cache discounts that can reduce effective cost for some usage patterns.
Is Fable 5 better than Opus 4.8?
Artificial Analysis data cited in the source says Fable 5 scores higher overall, at 64.9 versus 61.4 on the Intelligence Index. The same source says the advantage is larger on some sub-benchmarks and weaker on others.
Is the productivity gain proven?
No controlled public study is cited in the source. The main productivity example is Stripe’s Ruby migration story, attributed through Anthropic, which the source treats as unaudited customer evidence.
What should enterprise buyers watch next?
Workload-specific evidence will matter most: repeatable results, baseline comparisons, and cost per completed task. Until then, the published record supports testing before broad rollout rather than treating the price premium as already proven.
Source: Thorsten Meyer AI
