See how ChatGPT, Perplexity and Google AI Overviews describe you today

Free AI Visibility Audit
Surge45°
OpenAI model°

GPT-5.1-Codex-Max

GPT-5.1-Codex-Max is OpenAI's agentic coding model, built for long-running software development work that spans many files and many steps rather than single-prompt code completion.

OpenAIVerified 24/09/2026Released 04/12/2025

01 / decision

Decision snapshot

Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 126 models we track, not against an absolute standard.

Capability

Not measured

No published benchmark scores yet

Input / 1M tokens

$1.25

field median $0.30

Output / 1M tokens

$10.00

field median $1.25

Context

400K

field median 524K

02 / overview

What GPT-5.1-Codex-Max is, and when to reach for it

Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.

What it is

The job it was built for, and the job it is not for.

It is a coding model in the Codex line, based on an updated version of the 5.1 reasoning stack and trained on agentic tasks, meaning work where the model plans, edits, runs and checks rather than answering once. The job it is built for is sustained development: refactors, migrations, multi-file changes and tasks that carry a lot of repository context. It is not a general assistant for writing, chat or research, and it is not the cheap option for high-volume routine calls.

When it arrived, and when to use it

Where it sits in its line, and when a sibling is the better pick.

GPT-5.1-Codex-Max was first seen in December 2025, sitting alongside the general-purpose 5.1 models as the coding-specialised branch of the family. Reach for it when a task runs long, touches a lot of code at once, or needs the model to keep working without a human re-priming it each turn. It is the wrong pick for short prompts, summarisation, customer-facing chat or anything where a smaller general model would do the same job for less.

How you reach it

The API, the apps it powers, and what its limits let you do.

It is reached through the OpenAI API and powers Codex, OpenAI's coding surface, so it turns up in the terminal, the IDE and the cloud task runner rather than as a chat destination. The context window is large enough to hold a substantial repository slice plus logs and diffs in one session, and the output ceiling is sized for the kind of long patches and full file rewrites agentic work produces. It takes text and images, so screenshots, design mocks and error captures can go in with the code.

Why it matters

What changes because this exists, or why it does not.

The practical change is duration. Work that previously had to be chopped into small supervised chunks, because the model ran out of room or lost the thread, can now be handed over as one long task with the surrounding context attached. That matters most to teams doing migrations and large refactors, where the cost was never the model, it was the human stitching the pieces back together.

Follows GPT-5.1-Codex. Superseded by GPT-5.2. See the whole line.

03 / evidence

How much of this is verified

Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.

No published benchmark scores for this model yet.

We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.

How we decide what counts as evidence

04 / ledger

Benchmark ledger

Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.

Nothing in the ledger yet.

Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.

How we decide what counts as evidence

05 / capability

Capability shape

Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.

No category scores to shape yet.

A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.

How we decide what counts as evidence

06 / cost

What it costs

List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.

GPT-5.1-Codex-Max API pricingSurge45°
ChargePriceUnitRead onSource
Input$1.25per 1M tokens2026-09-24Check
Output$10.00per 1M tokens2026-09-24Check

It is priced as a working coding model rather than a premium reasoning tier, with input notably cheaper than output, which is the usual shape when a model is expected to read a lot and write a lot. Against the wider field it sits in the mid range, so the real budget question is not the rate but the volume: agentic runs generate far more tokens per task than a single prompt does.

What a month costsSurge45°
WorkloadInput / monthOutput / monthCost
A small product team20M tokens5M tokens$75.00
A busy support assistant200M tokens40M tokens$650.00
A document pipeline1000M tokens100M tokens$2,250.00

List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.

07 / specs

Specifications

As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.

SpecificationSurge45°
Context window400,000 tokens
Maximum output128,000 tokens
Modalitiestext, image
Released04/12/2025
StatusCurrent
Catalogue identifieropenai/gpt-5.1-codex-max

08 / lineage

Lineage

What this model replaced, what replaced it, and what else its provider has in the field.

Also from OpenAI

09 / line

The line

Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.

Came before

← GPT-5.1-Codex

Came after

GPT-5.2 →
  1. 01GPT-4.114/04/2025
  2. 02GPT-4.1 Mini14/04/2025
  3. 03GPT-4.1 Nano14/04/2025
  4. 04GPT-507/08/2025
  5. 05GPT-5 Mini07/08/2025
  6. 06GPT-5 Nano07/08/2025
  7. 07GPT-5 Pro06/10/2025
  8. 08GPT-5.113/11/2025
  9. 09GPT-5.1-Codex13/11/2025
  10. 10GPT-5.1-Codex-Max04/12/2025
  11. 11GPT-5.210/12/2025
  12. 12GPT-5.2 Pro10/12/2025
  13. 13GPT-6 Astra04/09/2026
  14. 14GPT-6 Astra Pro04/09/2026
  15. 15GPT-6 Luna22/09/2026
  16. 16GPT-6 Luna Pro22/09/2026
  17. 17GPT-6 Sol22/09/2026
  18. 18GPT-6 Sol Pro22/09/2026

Ordered by release date and worked out from the naming, so a new member slots in as soon as its page exists. A retired model keeps its page and its place in the line.

10 / notes

Our notes

What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.

What it changes for you

A coding model does not answer buyer questions about your product, so it will not cite you directly. What it changes is the competitive floor, since your rivals can now ship integrations, migrations and feature work faster, and the content and documentation that AI answer engines read gets refreshed more often. If your developer docs and technical pages are stale, the gap between you and a faster-shipping competitor will show up in what gets quoted.

Where buyers meet this model

Most buyers meet this one as developers, not as searchers: through Codex in the terminal or IDE, or through the API wired into their own tooling. It does not sit behind a consumer chat app or an AI search surface, so it shapes what gets built more than what gets cited.

Change log

Nothing published here yet. Changes appear within hours of a provider announcing them.

11 / questions

Questions

The things people ask about this model, answered from what is on this page rather than from anywhere else.

Is GPT-5.1-Codex-Max a replacement for the general 5.1 models?
No. It is the coding-specialised branch of the family, trained on agentic development tasks. For chat, writing, research and general product work, the general-purpose 5.1 models remain the right pick.
Can it handle images?
Yes. It takes text and images, so screenshots, UI mocks and captured error states can be passed in alongside code.
What does agentic actually mean here?
It means the model is trained to plan, edit, run and check across many steps rather than return one answer to one prompt. In practice you hand it a task, not a question.
Where do I actually use it?
Through the OpenAI API, or through Codex in the terminal, the IDE and the cloud task runner. There is no consumer chat destination for it.
Surge45°

Is GPT-5.1-Codex-Max recommending you?

Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.

Free AI visibility audit Book a call