OpenAI Launches Agents API in Public Beta, Halts Pro Plan Sign-Ups Amid Capacity Strain

New API enables long-running automated tasks; internal figures show researchers burning up to $7,000 daily on inference

edit
By LineZotpaper
Published
Read Time2 min
OpenAI released its Agents API in public beta on Thursday, giving developers the backend tools to run AI agents unattended for days or longer, while simultaneously pausing new sign-ups for its $200-per-month Pro plan due to capacity pressure from GPT-6 Astra demand. The juxtaposition highlights the company’s challenge of scaling infrastructure as agent usage accelerates.

OpenAI’s Agents API, the technology behind its Codex agent, is now available in public beta, allowing developers to build and run autonomous agents without managing their own orchestration layer. The API handles context compaction, letting tasks stretch beyond a single context window by compressing earlier work, and can invoke tools on demand or dispatch subagents to process parts of a job in parallel. Developers can execute agents inside OpenAI’s sandbox or on their own infrastructure.

While the move lowers the barrier for creating long-running agents, it also lowers friction for consuming inference. OpenAI’s own research organization provides a vivid example of how fast usage can scale. In a report published September 6, the company said its researchers logged the equivalent of 3.1 agent-workdays for every human workday by mid-August. The median researcher spent more than $600 per day on inference at API prices, and the 90th percentile exceeded $7,000 per day. Before June, researchers were still logging more hours than their agents; by mid-August, agents were doing three times the work of their human counterparts.

The timing of the launch coincides with a decision to halt new Pro subscriptions. Thibault Sottiaux, engineering lead for Codex, said on X that Pro plans “put the most strain on our systems” and that OpenAI was adding capacity as quickly as possible. The two products are separate, but both draw from the same underlying inference infrastructure.

§

Analysis

Why This Matters

  • Developers can now build agents that run for days without building orchestration from scratch, potentially accelerating the adoption of autonomous AI workflows.
  • OpenAI’s own internal usage data shows agent compute consumption can explode — a single power user can generate orders of magnitude more inference than a traditional developer, raising questions about cost and resource allocation.
  • The simultaneous Pro plan pause signals that even OpenAI is struggling to keep up with demand for its most intensive AI workloads, which could limit how widely agent capabilities scale in the near term.

Background

OpenAI has been gradually expanding its API portfolio beyond simple chat completions. The Codex agent, launched earlier this year, demonstrated the ability to complete software engineering tasks. The new Agents API generalizes that capability for any developer workflow. The company’s September 6 research report provided a rare internal look at how agents are consuming compute inside the company. The report showed that usage accelerated sharply after June, with agents overtaking human work hours by mid-August.

Key Perspectives

Developers: Gain a ready-made orchestration layer for long-running agents without building their own system. The API charges only for models, tools, and hosted compute used, reducing upfront cost for experimentation. OpenAI: Faces a tension between pushing agent adoption and managing infrastructure costs. The Pro plan pause suggests capacity constraints are real, and agent workloads could exacerbate the problem if adoption spikes. Skeptics: Caution that the ease of running agents for hours or days could lead to massive inference bills, as OpenAI’s own researchers’ spending demonstrates. The reliability and error-handling of unattended agents over long durations also remain unproven.

What to Watch

  • Developer adoption of the Agents API and any early reports of cost overruns or surprising usage patterns.
  • OpenAI’s infrastructure expansion announcements, especially capacity for GPT-6 Astra and agent workloads.
  • Whether the Pro plan suspension is temporary or leads to pricing changes or tier restructuring.
  • Competitors’ responses — Google, Anthropic, and others may hasten similar agent orchestration offerings.

Sources

newspaper

Zotpaper

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.