Fable 5 Makes Up Just 11% of Anthropic's Own AI Spending

The Financial Times reports Fable 5 makes up just 11 percent of total spending on Anthropic AI tools, based on Ramp payment data.

The Financial Times reported on August 23, 2026, that Fable 5, Anthropic's flagship AI model, accounts for only 11 percent of total spending across Anthropic's own product lineup. Ramp collected the figure from payment data covering 70,000 US companies. It has barely moved since Fable 5 launched in June.

Businesses aren't avoiding Anthropic. They're avoiding its most expensive model.

What the Spending Data Shows

Companies are reserving Fable 5 for their hardest problems. Everything else gets routed to cheaper alternatives. Some of those alternatives are Anthropic's own older models. Others are open-weight models from Chinese competitors.

The broader financial picture still looks strong on the surface:

• Anthropic's annualized revenue reached $65 billion in July, up from $47 billion in May.

• That figure still falls short of the $80 billion some bullish investors had projected.

• Anthropic reports 6,000 customers now spending $100,000 or more annually.

• The company expects Q3 to be profitable, using the same accounting approach it used to declare Q2 profitable.

OpenAI's numbers moved in the opposite direction over the same stretch. Its annualized revenue jumped 35 percent quarter to date to more than $40 billion. The FT ties that jump directly to GPT-5.6's July launch, after the model cleared US government review.

Why Fable 5's Launch Never Fully Recovered

Fable 5 didn't launch quietly. Its cyberattack capabilities triggered a temporary export restriction from the Trump administration, detailed in Claude Fable 5's shifting access terms. Anthropic later had that restriction lifted.

Anthropic expected adoption to accelerate once the political noise cleared. The Ramp data suggests that never happened. Fable 5's adoption curve is lagging behind Anthropic's own previous flagship launches, not catching up to them.

Accel partner Miles Clements, whose firm has invested $1 billion in Anthropic, offered a blunt read on the shift: "Most people don't need to operate at the frontier." That view runs counter to how frontier labs have priced their products so far.

Why This Matters Beyond One Company's Sales Numbers

If this pattern holds, it challenges the core business model every frontier AI lab has run on so far. Release a bigger, more expensive model. Count on customers upgrading to it. Claude Opus 5 launched as a mid-tier alternative to Fable 5 just weeks after this data started circulating, showing Anthropic already adjusting its own strategy around this exact pressure.

Google DeepMind's director of AGI economics, Alex Imas, offered a more measured read. He framed Anthropic's real concern as total spend across all its models combined, not Fable 5 in isolation. That's a fair point. It doesn't erase what the data shows about how businesses are actually allocating AI budgets right now.

Why This Matters for Businesses Buying AI

This data point is directly useful for procurement decisions happening right now. Most business tasks apparently don't need the most expensive, most capable model available. Cheaper, slightly older models are handling the bulk of real work at 70,000 companies. The flagship model gets reserved for a narrow slice of genuinely hard problems.

Businesses building their own AI spending strategy should treat that pattern as a benchmark, not an outlier. Defaulting every task to the priciest available model likely wastes budget. A tiered approach, frontier model for hard problems and a cheaper model for everything else, uses that budget more efficiently. Designing that kind of tiered routing into a real workflow is exactly the kind of problem agentic AI development is built to solve.

Key Takeaways

• Fable 5 accounts for just 11 percent of total spending on Anthropic's AI tools, based on Ramp data from 70,000 companies.

• Anthropic's annualized revenue hit $65 billion in July, below the $80 billion some investors expected.

• OpenAI's revenue grew 35 percent over the same period, fueled by GPT-5.6's cheaper pricing.

• Businesses are routing routine work to cheaper models, saving flagship models for the hardest tasks only.

• The pattern raises real questions about the AI industry's bet that bigger, pricier models keep driving revenue growth.

What Comes Next

Anthropic is reportedly moving toward what could be the largest public offering on record. That makes this spending pattern more than an academic curiosity. Businesses and investors should watch whether Q3 numbers show Fable 5 adoption catching up, or whether 11 percent becomes the new normal for what a flagship model actually earns.