OpenAI revolutionizes ChatGPT billing: no more paying without results

For the time being, major artificial intelligence providers use per-token billing. Thus, a customer purchases a certain number of tokens, enabling them to perform a certain number of actions. Important point: this mechanism by no means guarantees conclusive results.Some companies have thus spent staggering amounts on tokens to be able to use artificial intelligence models. For example, Microsoft canceled most of its Claude Code licenses four months after strongly encouraging its developers to use Anthropic’s agentic AI tools daily. The reason: the tokens provided by Anthropic quickly became expensive, without producing the expected results. OpenAI is aware of this problem. The company has also just made a significant change to its billing system.

OpenAI now charges some companies based on outcomes

As reported by the American media outlet The Information, OpenAI has introduced an outcome-based billing system. Thus, regardless of the number of tokens used by a client, it is the work actually performed that determines the bill. Unsuccessful attempts are not taken into account for billing purposes, with the associated costs being borne directly by OpenAI. At present, this outcome-based pricing system applies only to a few major OpenAI clients. We know neither the terms nor the prices.

Of course, an outcome-based pricing system is difficult to implement on a mass-market scale, given the substantial number of cases that would need to be studied. However, it represents a genuine turning point in the artificial intelligence sector.According to Keith Kirkpatrick, an analyst and journalist specializing in enterprise technologies: “ Outcome-based pricing is becoming the market standard”.

Why this change in the way some customers are billed?

The case of Microsoft, which canceled its Claude Code licenses after consuming a substantial number of tokens without achieving conclusive results, illustrates the problem on its own. Some companies burn through massive amounts of tokens without achieving conclusive results. Meanwhile, other companies see the problem and are curbing their AI investments. A company may also be unable to forecast the number of tokens it will need to successfully complete a project and may prefer to postpone it for fear of being faced with an exorbitant bill. Outcome-based billing thus provides a safer framework and a promise of completed work.

It nevertheless remains to be defined what constitutes work successfully and properly completed. Over the long term, will a client be able to assess that the work requested has not been carried out as expected, even as tokens continue to be consumed day after day? Precisely, the undisclosed terms of this type of contract should answer this question.

👉🏻 Follow the latest tech news in real time: add 01net to your feeds on Google, and subscribe to our WhatsApp channel.

Source: The Information