GPT 5.3 Codex
Neural Network
GPT-5.3-Codex is an OpenAI agentic model for code: development, refactoring, and debugging in a single workflow.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 400 000 tokens
- Max answer length
- 128 000 tokens
- Reasoning effort
- none, low, medium, high, extra high
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
Image prompt
(per 1K tokens)
import OpenAI from 'openai';
const openai = new OpenAI({
apiKey: '<your bothub access token>',
baseURL: 'https://openai.bothub.chat/v1'
});
// Sync - Text generation
async function main() {
const chatCompletion = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'gpt-5.3-codex',
});
}
// Async - Text generation
async function main() {
const stream = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'gpt-5.3-codex',
stream: true
});
for await (const chunk of stream) {
const part: string | null = chunk.choices[0].delta?.content ?? null;
}
}
main();
How it works GPT 5.3 Codex?
GPT-5.3-Codex is an OpenAI agentic model for development: it combines the engineering level of the Codex line for code work with the broader reasoning and professional knowledge of GPT-5.2. Unlike GPT-5.2-Codex, its strengths are not limited to programming, so the model handles long tasks where, besides code, you need to understand the subject area. In BotHub, it is accessible from Russia without a VPN or foreign card; payment is in rubles based on actual token usage, and the balance does not expire. Over 250 models are available in one window: you can compare answers and switch, while a unified OpenAI-compatible API allows you to change models in your project without rewriting the integration. Requests are encrypted, and for companies, there is a contract, invoice, and admin panel with limits. Use this model when you need to refactor a bloated service, understand an unfamiliar repository and find the cause of a bug, run a review of changes before merging branches, cover a module with tests, or complete a multi-step task with edits in several files at once.Frequently asked questions about GPT 5.3 Codex
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
The model is focused on working with code: writing functions, refactoring, debugging, analyzing third-party repositories, and code reviews. A 400,000-token context allows you to keep large projects in work entirely. It is also suitable for reasoning, STEM tasks, and analyzing documents and images, such as interface screenshots.
Up to 128,000 tokens in a single response — this is a full module, a large code file, or detailed technical documentation. If the task is larger, break it down into steps: the 400,000-token context allows you to keep the entire previous dialogue and not lose the thread.
Yes, the reasoning mode before answering is confirmed in our data. The model first breaks down the task into steps, checks hypotheses, and only then provides the result. This significantly helps with debugging, architectural decisions, and mathematics, where a correct answer is more important than a fast one.
Yes, both function calling and structured JSON output are supported. The model connects to your tools, databases, and external services, and returns the response in a predictable schema. In BotHub, this works via a unified OpenAI-compatible API, so you can change models without rewriting the integration.
Images and documents are accepted as input: error screenshots, diagrams, PDFs with technical specifications. There is no mention of audio or video in our data. If you need to work with sound or video clips, specialized models are available in the same BotHub window — switching takes a couple of seconds.
It is better to check with your own task: the language of prompts and responses is not described in our data, so we will not promise specifics. The service itself is fully Russian-speaking — interface, support, and payment with Russian cards, without VPN or foreign cards. A couple of queries will show the result more accurately than any description.