ChatGPT Pro parallel agents review: the work splits, the checking doesn't

What changed, in one paragraph?

On October 16, 2026, OpenAI updated GPT-6 Pro, the top model option in ChatGPT's Chat tab. OpenAI's release note lists three changes: faster responses, better use of plugins and tools, and multiple agents that can work on different parts of a task in parallel. You follow their progress in the conversation, and when a part is done you can open it and see what that agent did. It's rolling out to eligible Pro users. In company workspaces, it depends on how the workspace is set up.

That's the whole announcement. OpenAI didn't say how much faster, how many agents a task can use, or how it decides where to split the work. So what follows is about what the feature is for and what it costs you, not about benchmarks nobody has published.

Our short answer: if you already pay for Pro and you regularly ask ChatGPT for research or reports that break into separate parts, this makes those jobs quicker and easier to check. If you're on Plus and wondering whether to upgrade for this, probably not yet.

What is it actually good for?

Work that splits cleanly. Think of a question like "compare these four suppliers on price, minimum order, lead time and returns." One agent can work on each supplier while you watch, and you get four pieces you can open on their own. The same goes for a report with separate sections, a review of several documents, or checking one claim against several sources.

It's less useful when every step depends on the one before. If the answer to part two changes what you'd ask in part three, running them at the same time just means two of them guessed.

The bit we like most is the second half of the release note: you can open each completed task and see what that agent did. That's the difference between "here's a confident summary" and "here's where each piece came from." When a supplier's lead time looks wrong, you can open that agent's piece and see what it relied on, instead of rereading the whole answer and wondering.

Who gets it, and what does it cost?

Pro is now three monthly plans: Pro 100 at $100 a month, Pro 200 at $200 and Pro 500 at $500. Pro is billed monthly only, with no annual option. OpenAI's model page lists GPT-6 Pro, the model this update applies to, for Pro $100 and Pro $200 as well as Business and Enterprise. The parallel agents update itself is described as rolling out to eligible Pro users, so if you're on Business or Enterprise, check with whoever runs your workspace rather than assuming.

Two details matter for anyone budgeting. First, GPT-6 Pro has usage limits. OpenAI says plainly that a Pro subscription does not include unlimited use of the Pro model. On Pro $200, when you hit the weekly GPT-6 Pro limit, ChatGPT switches you to a lighter thinking model. On Pro $100, the Pro models share one weekly allowance. Second, on Business seats the Pro model is a small allowance: OpenAI's Business page lists 15 Pro messages a month on Standard seats and 50 a week on Premium.

Usage also depends on how much context, reasoning and tool use a task takes. A task split across several agents doing their own searching is not a small task. We'd expect big parallel jobs to use the weekly allowance faster than a single question, though OpenAI hasn't published how it counts them.

Where does it fall short?

It moves the work, it doesn't remove it. Five agents working at once still hand their work to one person. If you didn't have time to check one long answer before, you won't have more time to check five pieces now. The pieces are easier to check, which is real progress, but somebody still has to.

No published numbers. Faster by how much, more reliable tool use by what measure: the release note doesn't say. That isn't unusual for a rollout note, but it means "faster" is OpenAI's word, not a figure you can plan around.

Splitting can hide gaps. When a task is divided, the edges between parts are where things fall through. If one agent reads "price" as list price and another as price after discount, both pieces look fine on their own and the comparison is wrong. Read the joins, not just the parts.

Limits are the plan's limits. Heavy parallel work on the Pro model draws down a weekly allowance, and when it runs out you're moved to a different model mid-week. If a monthly report depends on it, know which week it's due.

It's for the Chat tab. This is GPT-6 Pro in Chat. OpenAI's other agent products, ChatGPT Work and dots, have their own allowances and their own rules.

How does it compare with Work and dots?

OpenAI now has three ways to hand ChatGPT a bigger job, and they suit different jobs.

Parallel agents in Chat are for a question you want answered now, in parts, inside one conversation.

ChatGPT Work is OpenAI's agent for longer tasks: research across connected apps and files, finished documents and spreadsheets, scheduled tasks. Its release note says you can follow progress, change direction and approve important actions as it works. It suits deliverables more than answers.

Dots are always-on agents with their own cloud computer that keep working between conversations and bring results back for review. We covered them in our OpenAI Dots review. They suit ongoing work, like watching for something or keeping a list current.

If what you want is "answer this hard, many-sided question properly," parallel agents in Chat is the closest fit. If you want something that turns into a document or keeps running next week, look at the other two first.

Should a business owner pay for it?

Pay for it if three things are true. You already ask ChatGPT for research or reports with separate parts at least weekly. You or someone on your team actually opens the pieces and checks them. And the weekly allowance on your plan covers the weeks that matter. In that case the update makes something you already do faster and easier to trust.

Don't upgrade for it if you're on Plus and mostly ask single questions, or if nobody will read the parts. A faster summary nobody checks is just a faster way to be wrong.

And if the job you're picturing is really a business process (the same research every Monday, the same comparison for every new supplier, an answer your customers depend on) a chat window may not be the right home for it at all. That's usually where a small, dedicated tool earns its keep: fixed sources, the same checks every time, and a named person who signs off. We build AI applications and agents like that. If you're weighing a chat subscription against something built for one job, talk to us.

Related work: Spin:Market.