OpenAI charges some large clients based on task completion results

According to The Information, OpenAI has recently started offering a new payment method to some major clients in the past few months, where clients only pay after the AI has actually completed the tasks.

The scenarios that have been implemented so far include customer service, where companies can pay based on the actual customer service tasks completed by the AI, rather than simply billing by Token, usage volume, or number of seats. OpenAI has not disclosed specific clients, unit prices, or the criteria for determining "task completion," and has declined to comment on this. This is not a new package available to all companies; the current public pricing on the official website for Enterprise still primarily charges based on quotas and Tokens, so this results-based billing seems more like a contract negotiated separately for certain major clients.

In January of this year, OpenAI CFO Sarah Friar suggested that future AI business models would adopt results-based billing. In July, OpenAI further stated that instead of focusing on the unit price per Token, it should be considered how much it actually costs for the AI to complete a task, and this line of thinking has now begun to enter real contracts.