TL;DR
Anthropic’s Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, twice the listed rate for Opus 4.8. Artificial Analysis benchmarks cited in the source material show a 5.7% Intelligence Index gain over Opus 4.8, while controlled public productivity studies have not been provided.
Anthropic’s Claude Fable 5, its most expensive active public model, is priced at $10 per million input tokens and $50 per million output tokens, while third-party benchmarks cited by the source material show a 5.7% Intelligence Index gain over Claude Opus 4.8, which costs half as much.
Anthropic pricing documents, its launch materials and product page are cited as confirming the $10/$50 per million-token rate for Fable 5. The source material says the price is 2.0 times Opus 4.8 and about 3.3 times Sonnet 4.6, making Fable 5 the highest-priced model in Anthropic’s active lineup.
Artificial Analysis benchmarks cited in the report put Fable 5 at 64.9 on its Intelligence Index, compared with 61.4 for Opus 4.8. The same benchmark set put Fable 5 at 1,932 Elo on GDPval-AA, compared with 1,890 for Opus 4.8.
Discounts apply in specific usage patterns. The source material says Fable 5 includes a 90% prompt-caching discount, a $5/$25 batch tier, and a calculated blended rate of $7.70 per million tokens under a 7:2:1 cache-hit-to-input-to-output mix. US-only inference is listed at a 1.1 times multiplier.
Price Comparison for AI Buyers
The comparison is relevant for enterprise buyers that evaluate premium AI models against cost, benchmark performance and measured productivity outcomes. In this case, the price is documented, while the public productivity evidence cited in the source material consists of one unaudited customer story in Anthropic’s launch materials.
The source material says Fable 5 is Anthropic’s highest-scoring public model on aggregators and that it set records in five of 10 sub-benchmarks. The aggregate comparison shows a procurement question for buyers: whether a workload benefits enough from those task-specific gains to justify a two-times price premium.
AI language model benchmark comparison
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The source material compares Claude Fable 5 with Claude Opus 4.8, because Opus 4.8 is listed at exactly half the price: $5 per million input tokens and $25 per million output tokens. Fable 5’s full Artificial Analysis Intelligence Index run is estimated at about $9,940, compared with about $4,970 for Opus 4.8.
The report also says Fable 5 and Mythos 5 use the same underlying weights and are priced the same, differing in safeguards. The source material says reports grouping them together on price are accurate on that basis.
“would otherwise have taken a whole team over two months by hand”
— Anthropic, citing Stripe
As an affiliate, we earn on qualifying purchases.
Public Evidence on Productivity
It is not yet clear whether Fable 5’s benchmark advantage translates into repeatable enterprise gains across broad knowledge-work tasks. The source material says there are zero controlled human-baseline productivity studies in the public evidence base and only one unaudited enterprise proof point.
The Stripe migration claim remains a claim Anthropic made, according to the source material. Public evidence has not established whether that example reflects a repeatable customer outcome, a workload-specific result, or an outcome that depends on internal tooling, task design and evaluation criteria.
AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Enterprise Evaluations Will Follow
The next evidence is likely to come from customer-side pilots, procurement reviews and independent evaluations comparing Fable 5 with cheaper models on real workloads. Buyers may measure task completion quality, latency, token mix and human time saved, in addition to leaderboard rankings.
Until controlled evidence is public, the confirmed points are limited: Fable 5 costs more, it scores higher on cited benchmarks, and the case for paying the premium depends on whether a buyer’s own workload aligns with Fable 5’s strongest benchmark areas.
OBD2 Scanner Reader for iOS & Android, Ai Diagnostic Tool for Car Buying & Repairs, No Subscription Fee, Lifetime Free Updates, Check & Clear Engine Codes, Real-Time Data, All 1996+
- Comprehensive Vehicle Diagnostics: Supports multiple tests and fault detection
- Real-Time Data Visualization: Displays data in clear, easy-to-read charts
- AI-Generated Car Health Reports: Provides quick, understandable health summaries
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?
The source material says Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens, before discounts such as prompt caching or batch pricing.
How much better is Fable 5 than Opus 4.8?
Artificial Analysis benchmarks cited in the report show Fable 5 scoring 64.9 versus 61.4 for Opus 4.8 on the Intelligence Index, a 5.7% aggregate gain.
Is there public proof that Fable 5 makes workers more productive?
The source material says there are no controlled public human-baseline studies. It identifies one unaudited customer story, attributed by Anthropic to Stripe, as the surviving public productivity proof point.
Does the higher price mean Fable 5 is a poor choice?
Not necessarily. The report says Fable 5 is Anthropic’s highest-scoring public model and performs better on some sub-benchmarks. The open question is whether a specific workload justifies the two-times premium.
Source: Thorsten Meyer AI
