Direct answer
Codex 5.3 means GPT-5.3-Codex, an OpenAI agentic coding model announced on 5 February 2026. The current API catalog still lists it, while GPT-5.1-Codex variants are deprecated. Check both the catalog and your live Codex model picker because availability differs by product and authentication method.
Summary
Do not treat an old launch article as a current availability page. The February 2026 model followed GPT-5.1-Codex and GPT-5.2-Codex with stronger agentic coding and computer-use claims. The current OpenAI catalog recommends the GPT-5.6 family for new complex reasoning and coding work. It still lists GPT-5.3-Codex and marks GPT-5.1-Codex variants deprecated. Update the client, select a model, and migrate an API integration as three separate tasks.
What is Codex 5.3?
The official API model ID for this informal name is `gpt-5.3-codex`. Searches for gpt 5.3 codex and gpt-5.3 codex refer to the same model. Codex is the coding-agent product and workflow; GPT-5.3-Codex is one model that has powered Codex and similar agentic coding environments. A Codex app or CLI version number is a separate thing.
Current status on 11 August 2026
- OpenAI's API page still lists `gpt-5.3-codex` for agentic coding tasks and documents low, medium, high, and xhigh reasoning effort.
- The all-model catalog places GPT-5.3-Codex under more models and marks GPT-5.1-Codex, GPT-5.1-Codex-Max, and GPT-5.1-Codex mini deprecated.
- OpenAI's current model guidance recommends GPT-5.6 Sol for new complex reasoning and coding, with Terra and Luna as other cost and throughput choices.
- The model available through an API key may differ from the models offered by a ChatGPT-authenticated Codex app, CLI, web, or IDE session.
- The user's live model picker and the current official catalog are the final checks for availability.
Check the current OpenAI model catalog before selecting or migrating.
What changed at the GPT-5.3-Codex launch?
OpenAI announced GPT-5.3-Codex on 5 February 2026. Its launch article said the model combined coding advances from GPT-5.2-Codex with reasoning and professional knowledge from GPT-5.2. OpenAI also reported a 25 percent speed improvement over GPT-5.2-Codex. These are launch claims from OpenAI, not an independent guarantee for every repository or task.
The launch emphasized long-running tasks, research, tool use, computer operation, and the ability to steer the agent while it works. OpenAI published results for coding, terminal, computer-use, and professional-work evaluations. Those results describe specific test setups and reasoning effort. Teams should reproduce relevant behavior on their own code, tools, policies, and acceptance tests.
Read OpenAI's original GPT-5.3-Codex announcement and evaluation notes.
GPT-5.3-Codex-Spark was a separate preview
OpenAI introduced GPT-5.3-Codex-Spark on 12 February 2026 as a smaller, latency-first research preview. At launch it was text-only, used a 128,000-token context window, and had separate access and rate limits. A Spark launch description should not be copied onto the main GPT-5.3-Codex model or treated as current availability.
GPT-5.3-Codex compared with GPT-5.1-Codex
The name gpt-5.1-codex can refer to the original GPT-5.1-Codex model, while GPT-5.1-Codex-Max was a later long-horizon variant. OpenAI's November 2025 Max announcement described compaction across context windows, stronger long-running work, and an xhigh reasoning option. It also said Max would replace GPT-5.1-Codex as the default in Codex surfaces at that time.
- GPT-5.1-Codex and its Max and mini variants are now marked deprecated in the current API catalog.
- GPT-5.3-Codex remains listed and supports a larger published context window and several reasoning-effort settings on its current model page.
- GPT-5.3-Codex's launch expanded the emphasis from software engineering toward computer use and broader professional work.
- A newer name does not prove a migration is safe. Prompt following, tool calls, diffs, tests, latency, token use, and safeguards can all change.
Read the original GPT-5.1-Codex-Max announcement for historical context.
Three meanings of update Codex
1. Update the Codex client
A client update changes the Codex CLI, desktop app, or IDE extension. It can add commands, fix bugs, change compatibility, and refresh model metadata. The Codex CLI inspected for this guide identifies itself as version 0.147.0 and exposes `codex update` plus `codex --version`. Read the help in your installed build and follow the distribution channel used by your team.
2. Select a different model
Changing the model picker or a `--model` option does not update the client. Confirm that the selected model is supported by the current authentication method and product surface. Run a small representative task and inspect the reported model before moving important work.
3. Migrate an API integration
An API migration changes a model ID or alias in code, configuration, or an environment setting. It may also require parameter, tool, context, pricing, rate-limit, or error-handling changes. Use the current model page and migration guidance, then validate behavior before sending production traffic.
Safe model migration checklist
- Inventory every hardcoded model ID, alias, profile, environment setting, fallback, test fixture, dashboard, and runbook.
- Check the current official model page, deprecation status, endpoint support, tools, context limits, reasoning settings, rate limits, and prices.
- Choose a supported target for the actual workload and authentication path instead of copying an old tutorial's model name.
- Create a fixed evaluation set from real coding tasks, reviews, tool calls, failures, and policy-sensitive cases.
- Run the old and candidate models with controlled prompts, tools, permissions, reasoning effort, and acceptance checks.
- Compare correctness, tests, diff size, regressions, tool behavior, latency, tokens, cost, refusals, and human review effort.
- Move a small share of reversible work, monitor errors and outcomes, and keep a documented rollback path.
- Update setup guides, examples, screenshots, support notes, and alerts only after the production choice is verified.
Use OpenAI's current model guidance when choosing a migration target.
What current comparison pages emphasize
Current competitor pages emphasize benchmark tables, speed, context, reasoning levels, cost, how-to steps, and head-to-head judgments. Artificial Analysis provides a direct GPT-5.3-Codex and GPT-5.1-Codex comparison. Launch coverage from TechRadar summarizes OpenAI's February announcement. This guide adds the missing time boundary: launch claims, API catalog status, subscription-surface availability, and local client versions must be checked separately.
Review the current Artificial Analysis comparison as third-party coverage.
Start with the broader explanation of what OpenAI Codex is.
Use the Codex installation guide when the client is not set up yet.
Review Codex usage and pricing separately from model capability.
Frequently asked questions
Do not infer current status from the February 2026 launch date. As of 11 August 2026, the API catalog still lists GPT-5.3-Codex, but OpenAI's general model guidance recommends the newer GPT-5.6 family for complex reasoning and coding. Check the live picker for the Codex surface and authentication method you use.
Usually. Codex 5.3 is common shorthand, while GPT-5.3-Codex is the official model name and `gpt-5.3-codex` is the API ID. A Codex client version such as 0.147.0 is separate and does not identify the active model.
Not necessarily. A client update installs a newer CLI build. The active model still depends on the picker, command option, configuration, authentication method, and current service availability. Verify the client version and selected model separately.