Corporate users are prioritizing cost efficiency over maximum model intelligence when selecting AI tools for enterprise deployment. Spending on Anthropic's top-tier model, Fable 5, has plateaued at 11% of overall outlay according to data from payments group Ramp. Concurrently, open-weight models captured 62% of tokens on Vercel AI Gateway in August, a significant increase from 28.4% in June. Fable 5 accounts for 6% of purchased tokens and costs twice as much as Claude Opus 5.
A forced withdrawal of Fable 5 in June damaged its pricing power and encouraged customers to qualify alternative models to reduce switching costs. Anthropic reports that the newer Opus 5 is within 0.5% of Fable 5's performance on CursorBench 3.2 while costing half as much per task. Despite the shift toward cheaper models, Anthropic's annualised revenue rose to $65 billion in July from $47 billion in May.
Key sources
- SOURCE@art_zucker“Anthropic’s best AI model struggles to attract users as cheaper tools thrive”x.com
- SUPPORT@code_star“You solve a problem, you pin the model, you don’t have a reason to upgrade it, it’s working!”x.com
- SUPPORT@rohanpaul_ai“Closed models already became the smaller piece.”x.com
- SUPPORT@rohanpaul_ai“max-effort Opus 5 came within 0.5% of Fable 5 on CursorBench 3.2 at half the cost per task”x.com
- SUPPORT@mathemagic1an“Marginal value add for Fable 5 is not worth it except in select cases that make up a minority of total usage by volume”x.com
- SUPPORT@maxforai“对于大量真实工作，模型能力到了某个水平以后，继续提升 10%、20%，用户的体验可能没差多少，但价格差距却非常明显”x.com
- SOURCE@burggrabenh“Spending on Fable 5, Anthropic’s largest and priciest model, has plateaued at only about 11 per cent of overall outlay on the company’s tools”x.com
- SUPPORT@burggrabenh“revenue last month hit $65bn on an annualised basis, up from $47bn in May”x.com