Home / Guides / Codex Max rates
Codex Max rates, and what replaced it
Codex Max, the model OpenAI named GPT-5.1-Codex-Max, cost $1.25 per million input tokens, $0.125 for cached input and $10 for output, with a 400,000-token context window. OpenAI shut it down on 23 July 2026 and named GPT-5.6 Sol as its substitute. Codex now runs on GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna, and the closest in price are GPT-6.1 Sol and GPT-6 Sol, both $2 and $10.
You may also be looking for one of two other things. Max is now a reasoning effort in the Codex model picker. And the plan with the most Codex usage is ChatGPT Pro, at up to $500 a month; there is no Codex Max plan.
Checked 3 November 2026 against OpenAI's GPT-5.1-Codex-Max page, its deprecations list and Codex pricing · By Soren Starck
| Model | API input | Cached | Output | Credits: input / cached / output | Plus, messages per 5 hours |
|---|---|---|---|---|---|
| GPT-5.1-Codex-Max (shut down) | $1.25 | $0.125 | $10 | Not listed | Not listed |
| GPT-6 Astra | $10 | $1 | $50 | 250 / 25 / 1,250 | 5 to 45 |
| GPT-6.1 Sol | $2 | $0.10 | $10 | 50 / 2.5 / 250 | 15 to 160 |
| GPT-6.1 Terra | $1 | $0.05 | $5 | 25 / 1.25 / 125 | 35 to 285 |
| GPT-6 Sol | $2 | $0.20 | $10 | 50 / 5 / 250 | 15 to 150 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 | 2.5 / 0.25 / 12.5 | 350 to 3,000 |
| GPT-5.6 Sol | $4 | $0.40 | $20 | 100 / 10 / 500 | Not listed |
API prices and credits are per million tokens at Standard speed. Message counts are OpenAI's estimates for local tasks, not fixed limits. OpenAI says API prices are separate from subscription usage and should not be used to estimate included tasks. GPT-5.6 Sol's API price is promotional until 30 November 2026 (read on 6 November 2026).
SessionWatcherWatch your Codex limits
whichever model you run.
Native app for the Mac menu bar and the Windows taskbar. Track Claude, Codex and 20 more tools' usage, costs, and rate limits in real time.
“Fast, simple, and does exactly what it should. Definitely worth it.”
@nicojerome on GitHub
macOS 14+. Free for 7 days, no credit card.
On Windows? Codex usage in the Windows taskbar: the same limits and alerts on Windows 10 and 11.
Can you still use Codex Max?
Not from OpenAI. Its model page marks GPT-5.1-Codex-Max as shut down, and its deprecations list says access ended on 23 July 2026, after a notice on 22 April. The Codex models page no longer mentions it. If a script or config file still names gpt-5.1-codex-max, change it.
You may still see the name elsewhere. Cursor's price table keeps a GPT-5.1 Codex Max row at the same rates, hidden by default; we did not test whether a request goes through.
What replaced Codex Max?
It depends on how you run Codex. On the API, OpenAI's substitute is gpt-5.6-sol, at $4 and $20 per million tokens. With ChatGPT sign-in, pick from the current models with /model or codex -m. Codex Max was described as optimized for long-running tasks, and OpenAI now suggests GPT-6.1 Sol for repeated, long-running work, so codex -m gpt-6.1-sol is the natural swap. Its output price matches Codex Max at $10, its input costs $2 instead of $1.25, its cached input is cheaper at $0.10, and it holds 500,000 tokens by default in Codex against 400,000. How that window works is in Codex context window.
What does one task cost on each model?
Take a mid-sized agent task with 40k new input tokens, 200k read from the cache and 20k of output, billed at the API prices above:
| Model | The same task |
|---|---|
| GPT-5.1-Codex-Max (shut down) | $0.28 |
| GPT-6 Astra | $1.60 |
| GPT-6.1 Sol | $0.30 |
| GPT-6.1 Terra | $0.15 |
| GPT-6 Sol | $0.32 |
| GPT-6 Luna | $0.02 |
| GPT-5.6 Sol | $0.64 |
Our arithmetic, not an OpenAI figure; the token counts are an example and cache writes are left out. The task stays under 272k input tokens: above that, OpenAI bills the whole request at its long-context prices, double the input and cached rates and 1.5 times the output (OpenAI's pricing page, re-read on 6 November 2026). It applies when you pay per token with an API key. A subscription draws on your plan's limits instead.
What does Max mean in Codex now?
Max is a reasoning effort, not a model or a plan. OpenAI says it gives the selected model more time to reason about a single task, and to use it for the hardest problems, when depth matters more than speed or usage: higher effort takes longer and uses more tokens. Ultra goes beyond a single-agent run, using subagents to handle separate parts of a task in parallel. OpenAI adds that most tasks need neither. In the CLI, run /model, choose a model, then More reasoning to see the options; GPT-6 Luna goes up to Max but not Ultra.
What is the highest Codex plan?
ChatGPT Pro, at $100, $200 or $500 a month, with Ultrafast mode on the $500 plan. There is no plan called Codex Max; Max is the name of Anthropic's Claude plan, which may be where the mix-up starts. OpenAI says Pro currently has no five-hour limit and that weekly limits may also apply. Plus at $20 has a five-hour limit, and the Codex 5-hour limit guide explains how both work. Whether Pro pays for your own use is in is ChatGPT Pro worth it for Codex, and every plan and the full rate card are in how much does Codex cost.
Which other Codex models are gone?
Codex Max went in a batch, and more have followed. If an old name still fails in your setup, it is probably one of these:
| Model | Gone from | Date | Use instead |
|---|---|---|---|
| gpt-5-codex, gpt-5.1-codex, gpt-5.1-codex-max | OpenAI API | 23 July 2026 | gpt-5.6-sol |
| gpt-5.1-codex-mini | OpenAI API | 23 July 2026 | gpt-5.6-terra |
| gpt-5.4, gpt-5.4-mini | Codex with ChatGPT sign-in | 31 August 2026 | gpt-6-sol, gpt-6-luna |
| gpt-5.3-codex-spark | Desktop app, CLI and IDE extension | 14 September 2026 | Any model on your plan |
| gpt-5.5 | Codex with ChatGPT sign-in; API-key sessions keep it | 14 October 2026 | A model on your plan |
OpenAI also lists gpt-5.2 and gpt-5.3-codex as already deprecated in Codex with ChatGPT sign-in, without a date.
Seeing what your Codex work uses
/status in the CLI and the usage page at chatgpt.com/codex/settings/usage show your limits when you stop to look. Every way to read them is in how to check Codex usage.
SessionWatcher keeps each limit OpenAI reports, five-hour and weekly, in the Mac menu bar or the Windows taskbar, with the time to each reset, and warns you at 70% and 90% by default. Its Codex token usage view adds your local token totals, an estimated API-equivalent cost and the sessions that used the most, which shows what a change of model does to your own numbers.
SessionWatcherSee what each Codex session uses
before the limit tells you.
Native app for the Mac menu bar and the Windows taskbar. Track Claude, Codex and 20 more tools' usage, costs, and rate limits in real time.
“Fast, simple, and does exactly what it should. Definitely worth it.”
@nicojerome on GitHub
macOS 14+. Free for 7 days, no credit card.
Frequently asked questions
Is Codex Max still available?
Not from OpenAI’s API. OpenAI lists GPT-5.1-Codex-Max as shut down: it announced the deprecation on 22 April 2026 and ended access on 23 July 2026. Codex’s own model page no longer mentions it. Cursor’s price table still lists a GPT-5.1 Codex Max row, hidden by default; we did not test whether a request goes through.
How much did GPT-5.1-Codex-Max cost?
$1.25 per million input tokens, $0.125 per million cached input tokens and $10 per million output tokens on the API. It had a 400,000-token context window and up to 128,000 output tokens.
What should I use instead of Codex Max?
OpenAI’s deprecations page names gpt-5.6-sol as the substitute on the API. In Codex with ChatGPT sign-in, the current models are GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna. OpenAI suggests GPT-6.1 Sol for repeated, long-running work, which is what Codex Max was built for. Run codex -m gpt-6.1-sol, and change any script or config file that still names the old model.
What does Max mean in the Codex model picker?
It is a reasoning effort, not a model or a plan. OpenAI says Max gives the selected model more time to reason about a single task, and to use it for the hardest problems, when depth matters more than speed or usage. In the CLI, run /model, choose a model, then More reasoning. OpenAI also says most tasks do not need Max or Ultra.
Does Codex have a Max plan?
No. The top individual plan is ChatGPT Pro, at $100, $200 or $500 a month. Pro currently has no five-hour limit, weekly limits may also apply, and Ultrafast mode comes with Pro $500. Max is the name of Anthropic’s Claude plan, which may be why the two get mixed up.
Sources
All read on 3 November 2026:
- GPT-5.1-Codex-Max: shut down, its prices, context window and max output.
- OpenAI API deprecations: the shutdown date and the substitute models.
- OpenAI API pricing: today's prices per million tokens and GPT-5.6 Sol's promotion.
- Codex models: the current models, the Max and Ultra efforts, and the retirements.
- Codex pricing: the plans, the five-hour estimates, Pro's limits and the credit rates.
- Cursor Models & Pricing: the GPT-5.1 Codex Max row.