How many GPT-6 Astra messages do you get in Codex and ChatGPT Work?
OpenAI estimates 5 to 45 GPT-6 Astra local messages every five hours on Plus and Standard Business in Codex and ChatGPT Work. Pro plans currently have no five-hour limit. Extra usage costs credits.
Checked against the lab's own pages on September 29, 2026. Limits change often, so the date matters.
OpenAI's estimate for GPT-6 Astra on Plus is 5 to 45 local messages every five hours in Codex and ChatGPT Work. Standard Business gets the same range. That's the smallest allowance of any GPT-6 model on the pricing page, and it's an estimate: OpenAI says these aren't fixed message limits.
- Plus, Astra
- 5-45 local messages per five hours
- Standard Business, Astra
- 5-45
- Plus, GPT-6.1 Sol
- 15-160
- Plus, GPT-6 Luna
- 350-3,000
- Pro
- no five-hour limit
- Astra credit rate
- 250 credits input, 1,250 credits output per 1M tokens
Why the range is so wide
Each message costs a different amount. OpenAI says a small script might use a fraction of your allowance while a long-running task that holds lots of context uses much more, and that model choice, reasoning, tool use and caching all move it. Cloud tasks can cost more than local messages, and the two share one allowance. Weekly limits may apply on top.
Astra is also the most expensive model per token. On credit-based plans it's 250 credits per million input tokens and 1,250 per million output, five times GPT-6.1 Sol's 50 and 250.
These estimates are not fixed message limits; check your usage dashboard for current limits and reset times.
What changes on Pro
Pro plans currently have no five-hour limit in Codex, according to the same page. Pro comes at $100, $200 or $500 a month, and only Pro $500 (plus eligible Enterprise and Edu plans) gets Astra Ultrafast. Ultrafast burns included usage at 8x the Standard rate, and Fast mode at 2.5x, so a fast Astra session eats a Plus-sized allowance in very few turns.
Codex and ChatGPT Work draw on the same usage. A long Work session in the browser comes out of the same pot as your Codex CLI runs.
When it resets, and what happens when you hit it
The window is five hours for Plus and Standard Business. Your usage dashboard shows the current limit and the reset time, and /status shows what's left inside a Codex CLI session. If you run out partway through a turn, OpenAI says the agent can finish that turn, subject to fair use limits.
After that, Plus and Pro users can buy credits instead of upgrading, and anyone can keep going with an API key at standard API rates. OpenAI's own advice for making it last is to pick a smaller model when it'll do: the models page suggests Astra for the hardest end-to-end work and GPT-6.1 Sol for repeated, long-running work when cost matters.
These numbers are as of September 29, 2026, the day GPT-6.1 Sol launched. Astra's estimate on Plus reads 5-45 and Sol's 15-160 on that date.