OpenAI launched the programming development model GPT-5.3-Codex on February 5, 2026. However, in the initial stage, the model is only available to ChatGPT paid subscribers in the OpenAI Codex, CLI, IDE plug-in and web version. Starting from the next day, some developers can call the model through the API platform.
Now the GPT-5.3-Codex model is fully available, and all developers can call the model through platforms such as OpenAI API platform or OpenRouter. The specific call prices of different platforms may vary.
The highlight of the GPT-5.3-Codex model is that it is not limited to code generation. It is currently OpenAI's strongest intelligent programming model. This model completely integrates the ultimate coding capabilities of GPT-5.2-Codex with the general reasoning and professional knowledge capabilities of GPT-5.2.
Key highlights include:
25% overall speedup: especially noticeable on long-term and multi-step agent tasks
Interactive guidance midway: You can intervene at any time during the task, modify the development direction, and supplement requirements without losing the context.
Multi-language real engineering test: SWE-Bench Pro covers Python, JavaScript, TypeScript, Java and other languages for the first time
Universal self-certification: The model directly participates in the debugging, training, running and scheduling deployment processes during its own training, that is, it improves itself.
Context window: supports up to 400K Tokens, suitable for very large code bases
New parameters: Support reasoning_effort (low/medium/high/xhigh) to finely control reasoning intensity
The official platform pricing is as follows:
Responses API pricing: $1.75 per million inputs, $14 per million outputs, $0.175 per million cached inputs (caching can significantly reduce duplicate context costs)
For more information, please visit the OpenAI API platform: https://developers.openai.com/api/docs/models/gpt-5.3-codex