ChatGPT's GPT-6 Astra Pro Can Now Run Several AI Agents at Once

OpenAI says the upgrade, which is rolling out to eligible Pro subscribers, also brings faster replies and better use of plugins and tools.

ChatGPT's GPT-6 Astra Pro can now split one request among several AI agents working at once, and Pro users can watch each one, OpenAI says.

The company announced the change Friday in a post from its @OpenAI account on X. OpenAI also says the model now answers more quickly and handles its plugins and tools better. Users can follow each agent as it goes and open a finished task to check its work, the company said. By Saturday, the post had nearly 1.6 million views on X.

The post says "Now available." OpenAI's ChatGPT release notes for the same day are more careful. They describe a gradual rollout limited to eligible Pro subscribers, with company-managed workspaces getting it only when their settings and eligibility allow.

Eligible means paying. The AI blog Kingy.ai, citing OpenAI's plan guide, said the model is offered on the $100 and $200 Pro tiers plus Business and Enterprise. It added that Plus subscribers don't get it in the regular Chat tab.

The pitch is that one prompt turns into several jobs at the same time. In a research request, one agent might chase pricing while another reads the documentation and a third looks for limits. Then the model stitches their findings into one answer.

Whether that saves time is still mostly OpenAI's claim. Kingy.ai tried it Friday on one eligible account, timing three agents in parallel against a single agent writing the same research memo. In the first pair, the lone agent beat the three-agent team by 10 seconds. The second time, the team won by 62 seconds.

The blog found no real accuracy edge either way. It said its four runs proved the feature works, not that it's reliably faster.

OpenAI's release notes don't put a number on "faster." Neither the notes nor the post say how many agents can run at once, or how a split-up task counts against a Pro plan's usage limits. Kingy.ai said ChatGPT didn't show a usage reading for each run, so it couldn't measure what delegating cost.

Even the name moved. Kingy.ai reported that the post's edit history shows its first version said "GPT-6 Pro" before it became "GPT-6 Astra Pro," and the release notes still use the shorter name. OpenAI's launch page for GPT-6 Astra, which the company calls the world's most intelligent model, promised Pro, Business and Enterprise users GPT-6 Astra Pro. Either way, it's the same model a Pro subscriber picks in the Chat tab.

Watching the agents work is the part OpenAI put up front. In Kingy.ai's test, three named agents started together and their status changed as each finished its part. But the activity panel kept saying "No outputs yet" after all three were done, and the finished work was sitting under a separate Subagents view.

Source: X, retrieved October 17, 2026. Other sources: releasebot.io, kingy.ai, openai.com.