OpenAI is experimenting with outcome-based pricing, charging customers only when its AI agents successfully complete assigned work, rather than billing for tokens or computing time used. This shift comes as the company prepares to launch Astra, its next-generation model designed specifically for AI agents that can work continuously for days or weeks.

What Is Outcome-Based AI Pricing?

Rather than the conventional approach of charging based on tokens (small units of text), queries, or computing time, outcome-based pricing ties the bill directly to results. If the AI agent doesn't complete the task, the customer doesn't pay. This model gives AI providers a strong incentive to build systems that actually deliver value, not just consume resources. Salesforce and other AI providers are also testing similar pricing approaches as the industry searches for new ways to monetize AI agents.

The shift reflects a broader evolution in how companies think about AI value. Instead of measuring success by how much computing power was used, outcome-based pricing measures it by whether the job got done. This could fundamentally change how enterprises budget for AI tools and how AI companies compete.

What Can Astra Do?

OpenAI's Astra model represents a significant step forward in AI agent capabilities. During a closed preview for VIP customers in early August, the company demonstrated several impressive abilities. According to tech journalist Alex Heath, who attended the preview, OpenAI CEO Sam Altman told attendees that Astra could be the first model where AI "invents new things in a way that matters".

Sam Altman

During the demonstration, OpenAI researchers showed Astra coordinating multiple agents to solve complex math proofs, working across desktop software at what Altman described as a "superhuman, very fast" pace. The model also handled practical business tasks including creating presentations, reviewing financial documents, and analyzing messy data sets.

"A new interface for continuous agents is coming very soon," stated Thibault Sottiaux, who leads OpenAI's combined product organization. "There isn't really an app, and the experience will be accessible from almost anywhere."

Thibault Sottiaux, Product Lead at OpenAI

However, OpenAI has delayed Astra's public launch. Internal testing revealed that the model could reach its highest "critical" cybersecurity capability threshold, meaning it could potentially autonomously find and exploit serious vulnerabilities. The company has not yet set a public rollout date.

How to Prepare for AI Agent Adoption in Your Organization

• Understand Outcome-Based Contracts: As AI pricing models shift from usage-based to outcome-based, review how your organization currently budgets for AI tools and prepare to negotiate contracts based on deliverables rather than compute consumption.
• Identify High-Impact Use Cases: Prioritize business processes where continuous AI agents could add the most value, such as financial analysis, data processing, or multi-step workflow automation that currently requires manual coordination.
• Plan for Security and Governance: As AI agents gain more autonomy and access to software tools, establish clear governance frameworks, access controls, and monitoring systems to ensure agents operate within intended boundaries.

How Does OpenAI's Growth Compare to Competitors?

OpenAI's revenue grew 18% quarter over quarter to $6.7 billion in the second quarter, according to recent reports. While this represents solid growth, it masks a concerning trend for the company. Anthropic, OpenAI's primary competitor, saw its sales double during the same period, signaling a significant shift in the competitive landscape.

The divergence highlights how quickly the AI race has shifted. Slower ChatGPT growth combined with the success of Claude Code, Anthropic's coding assistant, has put OpenAI on the back foot and forced a business and leadership overhaul. On the private market, Anthropic is now valued at $1.18 trillion compared to OpenAI's $878.32 billion valuation, according to Nasdaq Private Market data.

Both companies are pursuing public offerings. OpenAI has reportedly pushed its IPO to next year, while Anthropic may launch its offering as soon as next month or in early October. These moves signal that both companies are preparing for a new phase of competition and growth in the AI market.

The shift toward outcome-based pricing and the launch of more capable AI agents like Astra represent a maturation of the AI industry. As these systems become more autonomous and capable, the way companies charge for them and measure their value will likely continue to evolve. For enterprises, this means new opportunities to deploy AI agents on complex tasks, but also new considerations around pricing, security, and governance.

" }