TL;DR OpenAI upgraded GPT-6 Astra Pro in ChatGPT on October 16–17, 2026. For ChatGPT Pro subscribers, the model now delivers faster response times, stronger plugin and tool orchestration, and parallel multi-agent execution — several agents split one complex task and work on different parts at the same time. OpenAI also demoed "dots", always-on agents that autonomously triage late-night production incidents and hand off to Codex before an engineer even wakes up.

The short version

Until this update, a complex request in ChatGPT generally ran through a single agent that executed tool calls one after another. That serial approach works, but it bottlenecks work whenever a task spans several independent workstreams — you wait through each step in turn[2].

The new GPT-6 Astra Pro changes that. It can now spawn multiple agents that tackle different parts of a task in parallel, then report back their progress so you can inspect each finished piece individually[1][2].

Abstract AI neural network visualization, representing GPT-6 Astra Pro's parallel multi-agent execution

What actually changed

OpenAI announced the upgrade through its official channels and developer accounts, and the key changes for ChatGPT Pro users are[1][3]:

  • Faster responses — lower latency on Astra Pro across the board.
  • Stronger plugin and tool orchestration — the model routes more reliably across plugins, browsing, and connected tools.
  • Parallel multi-agent execution — the headline feature: one prompt can fan out to several agents working simultaneously.

The release drew significant attention on X, with the parallel-agents announcement pulling in 984,000+ views and 8,000+ likes within hours, according to community tracking[2].

How parallel agents work

Instead of a single agent grinding through a multi-part brief in sequence, GPT-6 Astra Pro now divides the work. Each agent handles its own component concurrently, and you can monitor every agent's live progress directly in the Chat tab, then open completed work artifacts one at a time to review the output[2][3].

In practice that means a request like "research three competitors, build a comparison table, and draft an outreach email" can run its research, table, and copy in parallel rather than back-to-back. The result is a shorter wall-clock time for compound tasks and a clearer view of what each sub-agent produced[2].

"dots": agents that work while you sleep

Alongside the Astra Pro upgrade, OpenAI Developers showcased "dots" — always-on personal agents used internally for on-call engineering. In the demo, an engineer leaves their dot on call overnight. When a 3 AM production alert fires, the dot autonomously triages the incident, investigates the root cause, and spins up a Codex coding thread — so the engineer can steer and approve the fix from the ChatGPT mobile app instead of opening a laptop[2].

OpenAI engineers also described workflows where Codex continuously scans repositories for tech debt and chips away at maintenance backlogs in the background[2].

Who gets it, and when

The upgrade targets ChatGPT Pro subscribers, where Astra Pro lives alongside the rest of the GPT-6 family[1][3]. If you're on Pro, select GPT-6 Astra Pro in the Chat tab and write prompts with clear, modular deliverables to trigger concurrent sub-agent execution[2].

Why it matters: This is a meaningful step toward agents that coordinate instead of just answering. For heavy ChatGPT Pro users, the practical win is faster multi-part work and the ability to audit what each agent actually did.

Sources

  1. OpenAI — official announcement on X — x.com/OpenAI
  2. AICoder — "OpenAI Upgrades GPT-6 Astra Pro with Parallel Multi-Agent Execution" — aicoder.com
  3. Tech & Business — "OpenAI adds parallel agents to GPT-6 Astra Pro in ChatGPT" — techandbusiness.org
  4. 36Kr — "ChatGPT有了AI小队，GPT-6 Astra Pro升级" — 36kr.com