Key Takeaways
- OpenAI now lets multiple agents work on a single task simultaneously
- Parallel execution doesn't speed up everything
- Usage limit consumption must be tested directly
- This week: count how many tasks can run at once
On October 16, 2026 (local time), OpenAI upgraded GPT-6 Astra Pro to let multiple AI agents split and simultaneously handle a single task. The announcement came via X, and the feature is available in the ChatGPT chat tab to some Pro plan subscribers, with availability varying by account.
What's changed
- Parallel processingBreaks down a single request into multiple parts, each handled simultaneously by a separate agent. Examples cited include reviewing large codebases and conducting complex research.
- Progress trackingUsers can view each agent's progress within the conversation screen, and once tasks are complete, open individual results to review the specific work performed.
- Speed and toolsOpenAI stated it has improved response speed and enhanced plugin and tool utilization capabilities.
For a corporate analysis report, for example, tasks like financial data research, competitor analysis, and market trend summaries can be distributed across multiple agents. In addition to the existing method where a single agent performs multiple steps sequentially, users can now choose to process independent tasks simultaneously. This update is not the release of a new AI model but an enhancement to the existing GPT-6 Astra Pro.
The approach of multiple AI agents collaborating is already used in AI coding tools like Anthropic's Claude Code and Cursor. What sets this update apart is that it brings this workflow to the general ChatGPT chat interface rather than a separate development environment.
Does running tasks in parallel speed up everything?
Not necessarily. If tasks depend on each other or multiple agents modify the same content, additional steps may be needed to merge results and resolve conflicts.
Usage consumption must also be assessed separately. Parallel execution can help complete tasks faster, but it doesn't necessarily mean reduced overall usage. OpenAI has not disclosed how parallel tasks factor into plan-specific usage limits, the maximum number of agents that can run in parallel, or the rollout timeline for Plus, Business, and Enterprise plans.
Implications for AX at Korean Enterprises
This feature operates within the chat interface rather than a development environment. Korean organizations should consider not only development teams but also departments like planning, strategy, and research that handle heavy research and report-writing workloads. Teams with many tasks that don't depend on each other's outputs—such as financial research, competitor analysis, and market trend summaries—have greater reason to evaluate parallel execution.
Start by checking plan eligibility. The feature is currently available to ChatGPT Pro plan users. Organizations using ChatGPT under Business or Enterprise plans should first confirm when the feature will become available on their plan before restructuring workflows around it.
Review units should also be reconsidered. Since completed tasks can be opened by individual agent, organizations adopting this feature should assign responsibility for reviewing each agent's output in advance, rather than checking only the final combined answer. Until clear usage benchmarks are released, it's safer to run the same task with a single agent versus parallel execution and directly compare usage consumption.
There's one thing to check this week: among your team's recurring research and reporting tasks, count how many can proceed simultaneously without waiting on each other's results. That number is the basis for evaluating this feature.
Source: OpenAI Upgrades 'GPT-6 Astra Pro' to Support Parallel Agent Execution