OpenAI Dots: always-on agents, their limits and real cost

OpenAI unveiled Dots at DevDay on 29 September 2026 — persistent cloud agents powered by GPT-6 Astra with their own computer, browser and memory. Chatting with a Dot is free of ChatGPT allowance, but Codex and Work tasks it launches draw on a five-hour window and cost about $0.04 per overflow credit.

DEV Community

Open source

OpenAI Dots: always-on agents, their limits and real cost

Dots run in ChatGPT, Slack, Teams and by voice, on Pro and Business Premium
Codex and Work tasks consume a five-hour window: 5–45 Astra messages on Plus
Stopping a Dot does not cancel already running background tasks or their spend
Pro excludes EEA, UK and Switzerland; local access needs the desktop app open

#OpenAI#ChatGPT#GPT-6#Codex

OpenAI Preps Always-On Agent "o" After Meta Muse Tops JP Morgan Benchmark

Real-SWE: Fable 5.1 beats GPT-6 Astra and Gemini 3.8 Flash on coding

Security

OpenAI Dots: always-on agents, their limits and real cost

OpenAI Preps Always-On Agent "o" After Meta Muse Tops JP Morgan Benchmark

Real-SWE: Fable 5.1 beats GPT-6 Astra and Gemini 3.8 Flash on coding

Palo Alto Networks launches always-on AI security testing on Claude Mythos and GPT-5.6-Cyber

OpenAI unveils Dots and GPT-6 Astra at DevDay 2026