OpenAI scraps plan to replace token pricing with flat fees

What's Happening

• OpenAI considered dropping token-based pricing for its next model, CEO Sam Altman said on the Cheeky Pint podcast.
• The company planned to let developers submit a prompt and receive a flat price for the work, similar to AWS billing.
• OpenAI abandoned the idea after customers pushed back against receiving a price without a breakdown of the costs.

Why It Matters

Developers rely on token counts to forecast costs and manage budgets for AI applications. An opaque pricing model removes their ability to predict expenses, especially for complex agentic workflows that consume variable amounts of compute.

The Context

• Developers track input and output tokens to manage costs and estimate spending as usage grows.
• Multistep agent workflows consume unpredictable amounts of compute, complicating budget forecasts under the current token system.

What They're Saying

• Altman told Stripe President John Collison that buyers rejected the opaque pricing, telling the company they need to know what they are paying for.
• Customers told Altman they would rather inflate the tokens and pay more to maintain visibility into the pricing structure.

What to Watch

• Altman acknowledged the limitations of token pricing but offered no clear alternative for measuring AI usage.

The Bottom Line

OpenAI will stick with token-based pricing because developers demand transparency over their compute costs.