Anthropic's Priciest AI Model Is Losing to Cheaper Rivals, Ramp Data Shows

Anthropic makes one of the most capable AI models on the market. Enterprise customers aren't buying it.

According to payments company Ramp, which analyzed spending data from 70,000 companies, Anthropic's flagship Claude Fable 5 accounts for only about 11% of total customer spending on Anthropic's tools, more than two months after its launch. That data was reported by the Financial Times and cited by KuCoin.

That's a problem for a company whose entire pitch has been frontier-level performance. Fable 5 runs between $10 and $50 per million tokens depending on input or output, according to Crypto Briefing. DeepSeek, the Chinese competitor, offers comparable performance for under $1 per million tokens. That's a different pricing universe.

Customers Are Routing Around the Flagship

AI startup Lindy moved its entire workload to DeepSeek, according to Crypto Briefing, purely because the math worked better. That's not an isolated call. Enterprises across the industry are adopting what's called model routing: send the easy stuff to cheap models, save the expensive ones for genuinely hard problems.

Anthropic's own cheaper model, Claude Opus 5, priced at $5 per million input tokens and $25 per million output tokens, has already overtaken Fable 5 in enterprise spending since its late-July launch, according to Ramp data cited by KuCoin. Anthropic is effectively getting outcompeted by its own product line.

Accel partner Miles Clements told the Financial Times, as relayed by KuCoin, that "most people don't need to operate at the frontier," and that the era when customers automatically chased the most powerful model "was not a sustainable era." That's a venture capitalist saying the premium-AI gold rush is cooling off, not a critic with an axe to grind.

The Political Wrinkle

Fable 5's launch in early June 2026 wasn't just a pricing story. According to KuCoin, the Trump administration forced Anthropic to withdraw the model at launch, citing national security concerns, and it wasn't relaunched until July 1 after government approval. Ramp's chief economist, Ara Kharazian, said data retention rules imposed by the administration also slowed the model's adoption.

Kharazian added that political factors are now playing a secondary role. Price and performance have become the dominant drivers of enterprise model choice. That's a specific, attributed claim: the government intervention mattered at launch, but it's not what's driving the ongoing shift toward cheaper models.

The Numbers Still Look Big, Just Not Big Enough

Anthropic isn't struggling by any normal business standard. The company has roughly 6,000 customers spending at least $100,000 a year, according to Ramp data cited by KuCoin. Dealroom reports Anthropic's annualized revenue hit $65 billion in July 2026, and Crypto Briefing put Claude's monthly active users at roughly 245 million by mid-2026, with the company capturing 41% of enterprise AI budgets as of May 2026, ahead of OpenAI.

But July's revenue came in below the most optimistic investor forecasts, which had expected the annualized run rate to top $80 billion, according to KuCoin. Average prices paid across Anthropic's models have dropped nearly 25% since mid-July, Crypto Briefing reported. Growth is real. It's just increasingly coming from cheaper products, not the flagship.

Meanwhile OpenAI, which Crypto Briefing says had a weak start to 2026, is closing the gap. KuCoin reports OpenAI's annualized revenue run rate rose 35% and topped $40 billion this quarter, driven by the July launch of GPT 5.6, priced well below Fable 5.

Why This Matters for a Potential IPO

Both Anthropic and OpenAI are reportedly preparing for possible public listings, according to Crypto Briefing. Dealroom's analysis put it plainly: model quality, usage, and revenue are separating from each other, and a frontier-model leaderboard isn't the same thing as a commercial leaderboard. Investors reportedly value Anthropic at $2 trillion or more ahead of any stock market debut, per KuCoin, a number that assumes the company can keep converting AI capability into revenue at scale.

If enterprise customers keep migrating to the cheapest model that gets the job done, that assumption gets harder to defend. Kharazian told the Financial Times that forecasting Anthropic's trajectory even a few months out is "extremely difficult" right now. Nobody quoted in this reporting, including Anthropic's own investors, seems confident they know how this settles before any IPO paperwork gets filed.