Anthropic's best model stalls while cheap tools thrive

Citing “people familiar with the matter,” the Financial Times has gathered several telling figures about the finances of the leading AI developers — recounted on his blog by Simon Willison.

Revenue is climbing

Anthropic's “annualized revenue” reached $65 billion as of the end of July — up from $47 billion in May. The company expects the third quarter to be profitable, by the same accounting method it earlier used to declare the second quarter profitable. Anthropic told investors it has 6,000 customers, each spending upward of $100,000 a year.

OpenAI is doing well too: its annualized quarterly revenue grew 35% and topped $40 billion. The push came from the July release of GPT 5.6 — after a sluggish start to the year.

Who pays for which models

The most interesting part is the Ramp AI index, which gauges the popularity of models from the payment data of 70,000 companies using Ramp corporate cards. The breakdown of spending on Anthropic's models for July 2026 looks like this:

• Opus 4.8 — 28.0%
• Sonnet 4.6 — 8.3%
• Fable 5 — 8.0%
• Opus 4.6 — 6.9%
• Sonnet 5 — 3.6%
• Opus 5 — 3.5%
• Opus 4.7 — 1.7%
• Sonnet 4.5 — 1.3%
• Haiku 4.5 — 1.0%
• Opus 4.5 — 0.7%

The picture, as Willison notes, looks plausible: Opus 5 came out only on July 24, so its modest share is understandable. But the low numbers for the pricey flagships bolster the FT piece's central point — Fable's high price has made the model less sought-after, while cheaper tools are steadily gaining an audience. In other words, having the “best” model is no guarantee that it's the one people will pay for.

Source: Simon Willison