OpenAI's price cut is a confession

Anthropic filed its confidential paperwork with regulators on June 1st. OpenAI, which had spent weeks insisting it felt no urgency to go public, filed its own a week later and announced the draft itself, rather than wait for the leak it assumed was coming. Two days after that, on June 10th, the Wall Street Journal reported that OpenAI was weighing drastic cuts to the prices it charges for tokens (the chunks of text its models read and bill against), to win back the developers and enterprises that have been drifting toward Anthropic. The sequence is the story. A company files to persuade the public markets that its business is worth the better part of a trillion dollars, then, almost in the same breath, contemplates the kind of price war that makes that case harder to argue.

Discounting is the move of the company that is behind on product, because price is the one lever a challenger can always pull and a leader would rather not touch. OpenAI knows it well, having spent the past year pulling it. When it launched GPT-5 last August, Sam Altman, OpenAI's boss, called the pricing the part he was happiest with, undercutting Anthropic's Claude on cost while matching it only roughly on the benchmarks that were supposed to settle the contest. GPT-4.1 had done the same to Anthropic's Sonnet tier a few months earlier. Read against that record, the latest cut is the frontier's defining company reaching, once again, for the weapon of the one that cannot quite hold it.