SAN FRANCISCO — When Anthropic released Fable 5, its most powerful model, the company’s sales team pitched it as the product that would justify the premium the market was placing on the company. Two months later, the spending data tells a different story. Ramp, a payments provider that tracks corporate software purchases, said Fable 5 accounted for only about 11 percent of spending across the 70,000 businesses it monitors that use Anthropic tools — a small share for a model marketed as the default choice for the company’s biggest customers.
The figure breaks the pattern that has governed enterprise AI purchases for the past two years. Corporate buyers historically opted for the strongest model a vendor offered, reasoning that extra capability justified extra cost. Fable 5’s reception suggests that logic has cracked. Analysts and investors who follow Anthropic said customers are migrating to cheaper alternatives for a straightforward pair of reasons: Fable 5’s price is high, and the company’s older models already handle the great majority of commercial workloads.
Ramp’s data, compiled from anonymized corporate card and procurement records, showed the shift is broad rather than confined to small startups. Large customers, the kind that sign multi-year contracts with AI vendors, are the ones with the most mature internal evaluations, and those evaluations increasingly conclude that the gap between Fable 5 and its predecessors does not justify a doubling of cost on certain workloads. Anthropic has declined to disclose Fable 5’s list pricing, but people familiar with the matter said the model carries a substantial premium over Claude Opus 5 for comparable usage.
The adoption data arrives as Anthropic prepares for an initial public offering that bankers and investors expect to be the largest in history. Market chatter has compared the deal to SpaceX’s $86 billion record, and some bankers have suggested Anthropic could exceed it, a bet that depends on investors accepting a valuation built on the assumption that enterprises will keep paying up for frontier models. Fable 5’s slow start puts a small but visible crack in that assumption.
Anthropic’s response has been to defend the model’s capabilities rather than its price. The company points to benchmark results showing Fable 5 leads on reasoning and coding tasks, and it has told customers that the model’s strengths show up in complex, high-stakes work — contract analysis, advanced engineering, long-horizon planning — where a small gain in accuracy outweighs the cost. Sales teams have also leaned on the model’s integration with the company’s broader platform, arguing that total cost of ownership matters more than per-token price.
Customers are listening, but slowly. Some large accounts have kept Fable 5 for a narrow set of workloads while shifting routine traffic to cheaper models, a pattern one executive described as “flagship for the crown jewels, workhorse for everything else.” That behavior, if it becomes the norm, would upend the economics of the frontier-model business, which relies on high prices for the strongest models to subsidize the rest of the product line.
The episode also sharpens the competitive picture. OpenAI has been cutting prices on its own frontier models and courting the same enterprise accounts; Google has bundled its Gemini models into cloud contracts at attractive rates. Anthropic’s premium pricing strategy assumed that enterprise buyers would pay for the best regardless of the gap. The Ramp data suggests buyers are now applying the discipline they use for every other software purchase: price, fit and proven reliability, in that order.
Anthropic’s analysts and investors say the migration is a pricing problem rather than a product problem, and they expect the company to respond with tiered packaging or usage-based discounts in coming quarters. For now, the numbers tell the story the marketing cannot. Two months after launch, the company’s most powerful model accounts for roughly one dollar in nine of what its customers spend. The valuation narrative is built on the promise that frontier models are indispensable; the payment data says they are optional.
The pattern is not unique to Anthropic. OpenAI has watched its own premium tiers mature faster than expected, and the company responded by cutting prices and bundling models into cloud subscriptions, a strategy that trades margin for usage. Google has pushed its Gemini lineup into enterprise deals at discounts designed to make switching expensive. The result is a market in which the marginal dollar is increasingly won on price and integration rather than on benchmark scores, and every vendor is being forced to explain why its flagship is worth the premium. Anthropic’s position is complicated by its timing: Fable 5 launched in the same quarter the company began meeting bankers to shape its IPO, and the model’s reception will feature in every roadshow presentation. A story that begins with “our best model is the market’s favorite” is easier to tell than “our best model is a third of our customers’ spend.”
What comes next is a test of the company’s flexibility. Anthropic has historically resisted discounting, arguing that its models’ safety and reliability justify list prices. That stance is already softening in negotiations with large customers, according to people familiar with the matter, and analysts expect the company to introduce tiered packaging that separates flagship capability from everyday throughput pricing. The deeper question for investors is whether frontier models can sustain premium pricing at all as open-source alternatives close the gap and buyers mature. Fable 5’s first two months suggest the market’s answer is more cautious than the valuation. Eleven percent of spend is not a failure — but for a model positioned as indispensable, it is a message, and Anthropic is hearing it.
