Loading...
OpenAI tests outcome-based pricing for AI
OpenAI's new pricing model is outcome-based

OpenAI tests outcome-based pricing for AI

Aug 31, 2026
04:31 pm

What's the story

OpenAI is said to be exploring a new pricing strategy for its artificial intelligence (AI) services. The company has started offering some of its major clients the option to pay only when its AI successfully completes assigned tasks. This marks a departure from the traditional usage-based pricing model that charges customers based on tokens, queries, or computing time.

Pricing innovation

Shift in pricing model

The shift to outcome-based pricing is not unique to OpenAI. Salesforce and other software companies are also testing similar models as they look for ways to monetize their AI agents.

This new approach gives providers an incentive to sell the result itself, rather than charging customers based on usage metrics like tokens or computing time.

Model preview

Astra next-gen model

The pricing shift comes as OpenAI gears up to launch Astra, its next-gen model focused on AI agents.

In a closed preview for VIP customers earlier this month, OpenAI CEO Sam Altman said Astra could be the first model where AI "invents new things in a way that matters."

The demonstration showed the model coordinating multiple agents to solve complex tasks at a superhuman pace.

ADVERTISEMENT

Model potential

Model to handle tasks like creating presentations, reviewing financials

Astra was shown to handle tasks like creating presentations, reviewing financials, and analyzing messy data.

Thibault Sottiaux, head of OpenAI's combined product organization, said a new interface for continuous agents is coming "very soon."

He added that there "isn't really an app" but the experience will be accessible from almost anywhere.

ADVERTISEMENT

Rollout plans

Astra's launch delayed over cybersecurity concerns

OpenAI is already using Astra internally but has not announced a public release date yet.

The model's launch was reportedly delayed after internal testing indicated it could reach its highest "critical" cybersecurity capability threshold, potentially allowing it to autonomously discover and exploit serious vulnerabilities.

ADVERTISEMENT