ThinkFacility

Plan limits

Codex usage limits for Plus: how many messages $20 a month buys

On Plus, OpenAI estimates 5-45 GPT-6 Astra, 15-150 GPT-6 Sol or 350-3,000 GPT-6 Luna local messages per five hours in Codex, with weekly limits on top. The estimates aren't fixed limits.

Checked against the lab's own pages on September 25, 2026. Limits change often, so the date matters.

Codex comes with every ChatGPT plan now, Free and Go included, and OpenAI's help article on using it with your plan says usage limits vary by plan. For the numbers it points you to the Codex pricing page. That page gives Plus, at $20 a month, a table of estimated local messages per five-hour window, and they're ranges, some of them wide.

Price
$20 /month
GPT-6 Astra
5-45 local messages per five hours
GPT-6 Sol
15-150
GPT-6 Luna
350-3,000
GPT-5.6 Sol
10-100
GPT-5.6 Terra
25-200

Why it's a range

The pricing page says the count depends on the model, how big and complex the task is, and whether it runs locally or in the cloud. A small script might cost a fraction of the allowance. A long session that holds a lot of context uses much more per message. OpenAI also calls these estimates, not fixed message limits, and sends you to the usage dashboard for the real figure and reset time.

Cloud chats on ChatGPT plans run on GPT-5.6 Sol and can use more of the allowance than local messages. Both come out of the same pool, and weekly limits may also apply over the five-hour ones.

These estimates are not fixed message limits

From Pricing | ChatGPT Learn

What shares the same allowance

I hadn't realized ChatGPT Work uses the same pricing, credits and usage limits as Codex, so a long Work task and a Codex session drain one pool. The help article adds ChatGPT for Excel and Workspace Agents to that shared allowance where they're on your plan. Image generations use included limits roughly 3-5x faster on average, and Fast mode burns credits at a higher rate.

When you hit the limit

Type /status in the Codex CLI, or open Settings, to see which allowance is spent and when it resets. If you hit it mid-turn, Codex can finish that turn. After that you can add credits (Plus users can buy them without changing plan), use a reset if you have one, upgrade, or wait. Support won't reset your limits. Pro is the upgrade path OpenAI names, with 5x or 20x Plus's usage.

These figures were read on September 25, 2026, and the pricing page notes GPT-5.5 leaves Codex on October 14, 2026.