Anthropic’s ambitious top-tier artificial intelligence model “Fable 5” is being ignored by enterprise customers. More than two months after launch, its share of corporate spending remains stuck at around 11%, losing ground to the company’s own cheaper models and competitors’ offerings. This is being interpreted as a signal that the competitive benchmark in the AI market is rapidly shifting from “peak performance” to “price-to-performance.”
The Financial Times reported on the 23rd (local time), citing data from payment services provider Ramp. An analysis of AI service spending across 70,000 U.S. companies found that enterprise spending on Fable 5 accounted for just 11% of total spending on Anthropic’s models more than two months after launch — essentially flat from the 11.4% recorded one month after release.
This shows the established pattern of enterprise users defaulting to the most powerful latest models is breaking down. Analysts and investors point to Fable 5’s high price as the primary culprit. Existing or relatively smaller models can handle most enterprise workloads, reducing the need to pay a premium for top-tier performance.
Fable 5 Loses Ground to Both In-House and Rival Models
Even within Anthropic, a cheaper model is outpacing Fable 5. “Opus 5,” released in late July, is a smaller model than Fable 5 but has already surpassed it in enterprise spending by offering strong performance at a lower price. According to Ramp data, Opus 5 costs roughly half as much as Fable 5, yet sometimes delivers better results in programming and knowledge-work evaluations.
Rival OpenAI is also staging a comeback on price competitiveness. GPT-5.6, released last month, is significantly cheaper than Fable 5 and has helped revive momentum that had slowed earlier this year. Ramp tracking data shows GPT-5.6 captured 25% of enterprise token usage and 23% of spending during the same period, overwhelming Fable 5.
Fable 5’s total revenue contribution was only about 75% of GPT-5.6’s. Given that its per-token price is twice as high, this suggests enterprises are shunning Fable 5 even more dramatically in terms of actual usage. Fable 5’s unit price is double that of Anthropic’s own Opus 4.8 and ten times that of Haiku 4.5.
Open-Source Models Add Pressure
The proliferation of open-source AI models is also narrowing Anthropic’s position. According to Vercel AI Gateway statistics, open-source models’ token share climbed from 11% in April to 29% in June, then surged to 62% by August. Yet they account for less than 4% of total enterprise AI spending.
If this trend continues, the business models of AI developers that have concentrated billions of dollars in development costs on training ever-larger and more sophisticated frontier models could fundamentally change.
Miles Clements, a partner at Accel — which has invested approximately $1 billion (~1.4 trillion won) in Anthropic — said “most people don’t need to use AI at the frontier level,” adding that the era when customers defaulted to only the most advanced models “was not a sustainable period.”
He noted that performance innovation remains important for achieving the long-term goals AI companies tout, such as curing diseases, and for attracting top research talent. However, he suggested top-tier models may increasingly serve as “showcase” demonstrations of technical prowess rather than drivers of actual usage.
Anthropic’s Growth Remains Strong Ahead of IPO
Fable 5’s weak demand is drawing attention as it coincides with Anthropic’s preparations for what could be the largest initial public offering in history. Investors are valuing Anthropic at more than $2 trillion, and the listing could begin as early as next month.
Anthropic’s annualized revenue in July reached $65 billion (~89.7 trillion won), falling short of the $80 billion (~110.4 trillion won) some investors had expected but growing substantially from $47 billion (~64.8 trillion won) in May. The company has grown revenue roughly sevenfold since the start of the year and posted its first adjusted operating profit in the second quarter. The number of customers spending more than $100,000 (~140 million won) annually has reached 6,000.
However, Ramp chief economist Ara Karazian noted, “Extrapolating previous trends would have suggested Anthropic would dominate the market, but with OpenAI’s latest model performing well and Fable underperforming expectations, the outcome has been the opposite.” He added that predicting Anthropic’s trajectory over the coming months is “nearly impossible.”
As enterprises begin selecting models based on price and performance for specific use cases — rather than applying the most powerful model to every task — the competitive landscape of the AI market is being reshaped. The symbolic value of frontier models may persist, but the models likely to drive actual revenue are increasingly those with practical price competitiveness.