GPT-6 Pro in ChatGPT now runs tasks across parallel agents
GPT-6 Pro in ChatGPT's Chat tab can now split tasks across multiple agents working simultaneously, with inline progress tracking for Pro subscribers.
On 16 October 2026, OpenAI updated GPT-6 Pro in ChatGPT’s Chat tab with parallel multi-agent execution. Where GPT-6 Pro previously worked through a task step by step, it can now break a task into parts and hand each one to a separate agent, all running at the same time. You can follow each agent’s progress inline within the conversation, and once they finish, you can open any completed task to see exactly what that agent did.
This is part of a bundle of improvements to GPT-6 Pro in the Chat tab: faster responses and better plugin and tool use are included alongside the multi-agent work. The multi-agent capability is the most structurally significant change, though, because it shifts the model from a single assistant working sequentially to something closer to a small coordinated team.
How parallel agents actually work
The idea is straightforward. Take a research task with three independent components: competitor analysis, pricing benchmarking, and regulatory review. Handled sequentially, each waits for the previous one to finish. With parallel agents, GPT-6 Pro delegates all three simultaneously and stitches the results together once they are done.
The caveat is equally straightforward. Parallelism only helps when the subtasks are genuinely independent. If step two depends on the output of step one, the agents still have to wait, and the benefit disappears. OpenAI has not published guidance on how the model decides which parts of a task can run in parallel or how many agents it will spin up for a given request.
Independent testing reported roughly a 9% speed improvement on average across parallel runs compared with sequential controls, though the variance was high and no clear accuracy advantage was found. Those numbers come from a small sample, so treat them as an early indication rather than a settled benchmark.
What this means for you
If you are a Pro subscriber, this is a meaningful quality-of-life upgrade for complex, multi-part tasks. Instead of watching one long response scroll past, you get visibility into what is happening at each stage, and the work finishes faster when the subtasks can genuinely be parallelised.
The practical value depends on what you use ChatGPT for. If your typical sessions involve straightforward back-and-forth questions, you will not notice much difference. If you regularly ask GPT-6 Pro to do things like draft a report while pulling data from several sources, or research a topic across multiple angles at once, parallel agents give you more transparency and, in the right conditions, faster results.
The inline progress tracking is a notable usability addition on its own. Knowing which agent is working on which part of a task, and being able to inspect its completed work, makes the process legible in a way that a single opaque response does not.
Who can access it
Parallel agents are part of GPT-6 Pro, powered by GPT-6 Astra. That model is available on specific plans only.
| Plan | Access |
|---|---|
| Free / Go | Not included |
| Plus | Not included |
| Pro 100 / Pro 200 | Included (subject to usage limit) |
| Business | Included (subject to workspace settings) |
| Enterprise | Included (if admin has enabled the model) |
To use it, select GPT-6 in the Chat tab and set the level to Pro. The rollout is gradual, so the feature may not appear on your account immediately even if your plan qualifies. Business and Enterprise customers should check their organisation’s workspace settings, as availability depends on admin configuration.
Worth noting: this is not a new model. The 16 October update is an incremental improvement to the existing GPT-6 Pro Chat experience. OpenAI made a broader GPT-6 rollout on 7 October 2026, when GPT-6 and an updated UI reached Plus, Pro, Business, and Enterprise globally. The parallel agent capability builds on that foundation rather than replacing it.
What OpenAI has not said
OpenAI has not published latency figures, quality benchmarks, a limit on how many agents can run in parallel, or an explanation of how parallel agent usage counts against a Pro plan’s allowance. Those are meaningful gaps for anyone trying to plan their usage or assess cost implications. It is also unclear which worker models the agents use, how automatic delegation decisions are made, or how consistent plugin reliability is across parallel runs.
Community reaction since the rollout has been mixed. Some users are genuinely enthusiastic about the multi-agent capability. Others are more focused on concerns about usage limits and the complexity of the tiered plan structure. Both reactions are reasonable given the limited transparency around how the feature consumes quota.
A related update on 17 October 2026 touched Codex and ChatGPT Work: subagents can now reuse context, cached context stays available for longer, and request routing has been adjusted to increase cache hit rates. Those changes sit adjacent to the Chat tab update but are aimed at a different part of the product.
The bigger picture
ChatGPT’s Chat tab has historically been the product’s most accessible surface, a place for conversation rather than agentic work. Adding parallel agents here, with inline visibility, is OpenAI moving agentic behaviour into the mainstream ChatGPT experience rather than keeping it in separate tools or API-only features.
Whether that translates into meaningfully better outcomes for complex tasks will depend on how well the delegation logic matures, how clearly OpenAI communicates usage costs, and whether the reliability holds across diverse task types. For now, it is a substantive addition to what GPT-6 Pro can do inside a standard conversation.