clauding.de
DE
• 2 min read AI-generated

Anthropic Is Making Serious Money – But the Expensive Top Models Sit Idle

Copy article as Markdown

Fresh numbers from around Anthropic show a big jump in revenue. The more telling part is what companies actually pay for: the Ramp index reveals that pricey Opus 5 barely gets used – while a model from May tops the list.

Anthropic Business Pricing
Featured image for "Anthropic Is Making Serious Money – But the Expensive Top Models Sit Idle"

Sometimes a single table says more than a whole earnings report. This week the Financial Times dropped a few numbers from around Anthropic, and Simon Willison paired them with a second source that interests me even more.

Money first. Anthropic’s annualized revenue hit roughly $65 billion in July. In May it was $47 billion. Q3 is expected to be profitable – by the same math that already declared Q2 profitable. And 6,000 customers spend at least $100,000 a year. OpenAI is growing too: over $40 billion, up 35 percent in the current quarter, boosted by the GPT-5.6 launch in July.

The number that sticks

What’s really revealing isn’t the revenue, it’s what people pay for. The Ramp AI Index reads the card statements of 70,000 companies and estimates from them which models are in use. Here’s Anthropic’s July breakdown:

  1. Opus 4.8: 28.0%
  2. Sonnet 4.6: 8.3%
  3. Fable 5: 8.0%
  4. Opus 4.6: 6.9%
  5. Sonnet 5: 3.6%
  6. Opus 5: 3.5%

Notice anything? The most-used model by a wide margin is Opus 4.8 – released back in late May. Brand-new Opus 5, out only on July 24th, scrapes together a measly 3.5 percent. And Fable 5, the priciest top model in the lineup, sits at 8 percent.

My take

The pattern is pretty clear: companies reach for the proven workhorse, not the most expensive newcomer. A model has to earn its spot, and that takes time. Opus 5 isn’t even a month old, and plenty of teams simply haven’t rewired their pipelines yet. Anyone who has tried Fable 5 quickly sees what the price does to the bill.

None of this is a problem for Anthropic – the revenue is there. But it shows how little the hype around each new flagship has to do with daily reality. Out in the field, the work runs on models that are good enough and stay affordable. Expensive and brand-new is a sales argument – not a usage one.

Sources: