OpenAI Brings GPT-6 Pro parallel agents to ChatGPT

OpenAI's GPT-6 Pro parallel agents turn ChatGPT's Pro mode into a coordinator. In release notes published on October 16, 2026, the company announced that the feature can now split a single request among several agents working in parallel inside the Chat tab. Users can follow each agent's progress in the conversation and open finished tasks to inspect what every agent did, according to OpenAI's ChatGPT release notes.

The update also brings faster responses and more effective use of plugins and tools when carrying out tasks, according to the release notes. OpenAI has published no latency measurements, quality benchmarks, agent count limits, or a formula for how multi-agent work counts against a Pro plan's usage allowance — a gap that independent coverage has been quick to point out.

How GPT-6 Pro parallel agents work

Instead of one assistant working through a request step by step, the model can now hand separate parts of the same task to different agents so they work at the same time. While the agents are working, their progress shows up in the conversation itself. When they finish, each completed task can be opened to check what each one has done.

The feature fits jobs with independent lanes and a clear final deliverable. Ask ChatGPT to assess an AI service, and one agent can verify pricing and plan rules while another examines technical behavior and a third lists limitations, then GPT-6 Pro combines the results into one answer. For a product comparison, each agent can work from the same extraction template so the results reconcile cleanly, according to hands-on testing by Kingy.ai.

The rollout is gradual and is reaching Pro plan users who meet the requirements. In managed workspaces, availability depends on workspace settings and rollout eligibility. The feature is not a new model: GPT-6 Pro in Chat is powered by GPT-6 Astra, and OpenAI's model guide lists GPT-6 Pro for the Pro $100, Pro $200, Business, and Enterprise plans. The release notes describe GPT-6 Pro parallel agents as one half of a wider October 16 update that also brought faster answers and improved plugin and tool use.

An independent test of GPT-6 Pro parallel agents: modest speed gains, no accuracy edge

Kingy.ai tested the feature on October 16 in an eligible ChatGPT account, running two sequential-versus-parallel research pairs with a frozen three-lane brief. The parallel runs averaged 4 minutes 25 seconds against 4 minutes 51 seconds for the sequential controls — about 9% faster — but the individual results varied sharply: the first parallel run was actually ten seconds slower than its control, and the second was 62 seconds faster. The test found no clear accuracy advantage for either mode, according to the outlet's published results.

What the test did find was an inspection surface that ordinary one-stream answers lacked. Chat showed three named agents starting together, updated their status as each lane progressed, and reported all three as finished. Each completed task opened into its own panel with the agent's full result, sources, qualifications, and unknowns, which made it possible to check whether the final synthesis had dropped a caveat. The common complaint about long-running agent work is not that it is slow but that it is silent, and visible per-agent progress addresses that complaint directly.

The test also exposed rough edges. The activity panel continued to say "No outputs yet" after all three agents had finished, even though their completed results were available under the subagents view. Chat did not expose a per-run usage reading, so the allowance cost of delegation could not be measured, according to the report.

What OpenAI left open

The announcement's gaps are central to the buying decision for Pro subscribers. OpenAI has not said how many agents can work at once, whether Chat delegates automatically or only on explicit instruction, which worker models the agents use, or how a parallel request counts against the plan's usage allowance. The post offers no latency figures or examples to support its claims that GPT-6 Pro responds faster and uses plugins and tools better.

There is also a naming wrinkle. OpenAI's X post currently calls the product "GPT-6 Astra Pro," though its edit history shows the first version said "GPT-6 Pro," while the Help Center consistently describes "GPT-6 Pro, powered by GPT-6 Astra." For readers, GPT-6 Pro is the Chat product label and Astra is the model behind it.

GPT-6 Pro parallel agents in Chat should not be confused with OpenAI's other multi-agent products. The always-on Dots agents, announced at DevDay 2026, run around the clock on cloud computers for eligible users — a different surface from parallel task workers inside a conversation. OpenAI's Agents API and Responses API expose related multi-agent patterns for developers, but their agent counts, controls, model choices, and token accounting are separate products and should not be treated as specifications for ChatGPT, according to Kingy.ai's analysis.

For the agent ecosystem, the release matters as a design statement as much as a feature. The visible-progress design turns a black box into something a user can interrupt — and each inspectable task record is a check on what the coordinator actually did with delegated authority. As GPT-6 Pro parallel agents move from announcement to everyday use, the open questions of agent limits, usage accounting, and default delegation rules will decide whether the feature becomes a habit or a demo. Similar platform moves, like DigitalOcean's agent learning services, show the industry converging on reviewable, accountable agent work rather than faster black boxes.

Sources: OpenAI ChatGPT release notes · Kingy.ai hands-on test · Runtime Wire · AI Brief