Anthropic’s most powerful and most expensive model, Fable 5, has not become a default enterprise choice more than two months after launch. Citing a Financial Times report, BlockTempo said payments platform Ramp analyzed spending data from more than 70,000 businesses and found that Fable 5 still accounted for only about 11% of total spending on Anthropic tools, while nearly 90% of budgets continued to flow to older, cheaper models.
Fable 5 remains stuck near an 11% share of enterprise spending
According to the report, the numbers break with the earlier pattern in which enterprise customers often gravitated toward the strongest model available.
Ramp’s sample showed that Fable 5 represented 11.4% of Anthropic spending when measured by dollars. Measured by token usage, or the amount of text the model actually processed, its share fell to 6%.
At the same time, Opus 5, a smaller and lower-priced model, has already overtaken Fable 5 in business spending since its late-July release.
OpenAI offers a contrasting case
The Financial Times also pointed to OpenAI as a comparison. After launching the clearly lower-priced GPT 5.6 in July, OpenAI’s annualized revenue for the quarter to date jumped 35% to more than $40 billion. The report defined annualized revenue as taking one month’s revenue and projecting it across 12 months.
That shift reversed a period in which OpenAI had trailed for much of the year.
Investor view: most customers do not need frontier performance
On why companies may not choose the most expensive model, the report cited Miles Clements, an Accel partner whose firm has invested nearly $1 billion in Anthropic. Clements told the Financial Times, 「Most people simply don’t need to operate at the performance frontier.」
He also said the phase in which customers insist on frontier models is 「not a durable long-term norm.」
Clements said intelligence breakthroughs still matter because they are tied to Anthropic’s larger promises, including claims around curing disease, and because they help the company attract top researchers. For most enterprise customers, though, the most advanced models are likely to become more of a showroom product than an everyday tool.
Lower-cost open-weight models add pressure
The Financial Times said low-cost open-weight models from China and other regions are also giving customers more alternatives. These are models whose internal weights are made available so outside users can deploy and adjust them on their own.
That has reduced the case for always paying for the most expensive option.
Anthropic’s business is still growing and has shown signs of profit
Even so, weaker-than-expected uptake for Fable 5 has not translated into a broader business slowdown. The report said Anthropic, led by Chief Executive Officer Dario Amodei, has grown revenue nearly sevenfold this year and posted positive adjusted operating profit for the first time in the second quarter, meaning its core business turned profitable after excluding some one-off costs.
Citing people familiar with the matter, the Financial Times said Anthropic told investors it could return to profit in the third quarter and currently has 6,000 large customers spending at least $100,000 a year.
IPO test now centers on whether growth can hold
The report added that it remains unclear whether that momentum will carry through to Anthropic’s IPO. Last month, the company’s revenue reached an annualized $65 billion, below the most optimistic investor expectation of $80 billion.
It also said Anthropic’s growth curve slowed in June after the Trump administration restricted Fable promotion, only stabilizing again in recent months.
Anthropic is now preparing for an IPO that outside investors expect could arrive as early as next month. Those investors estimate the company could be valued at more than $2 trillion, which would make it the largest initial public offering on record if that figure is achieved.
For frontier AI labs, the report framed this as a direct business-model test: after pouring research budgets into training ever-larger models, they may be finding that most customers simply want versions that are good enough and cheap enough.