# GPT 5.3 Codex

OpenAI's GPT-5.3 generation codex model. It builds on two generations of refinement since GPT-5.1, with measurable gains in code comprehension depth and multi-file reasoning.

- **Model ID:** `openai/gpt-5.3-codex`
- **Type:** chat
- **Providers:** openai, azure
- **Context window:** 400,000
- **Maximum output tokens:** 128,000
- **Pricing:** $1.75/1M input tokens, $14/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/gpt-5.3-codex

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'openai/gpt-5.3-codex',
  prompt: 'Why is the sky blue?'
})
```

## About

GPT 5.3 Codex became available on February 5, 2026 on AI Gateway as OpenAI's GPT-5.3 generation codex model. Two generations of refinement since GPT-5.1 codex translate to measurable gains in code comprehension depth and the ability to reason across files in large repositories.

The GPT-5.3 generation sharpens the codex agent loop (reading context, planning changes, writing code, running tests, and iterating) with better accuracy at each stage. Where earlier generations occasionally missed cross-file dependencies or produced partially correct refactors, GPT 5.3 Codex handles these patterns more consistently.

The context window of 400K tokens gives GPT 5.3 Codex comprehensive codebase understanding, letting it trace call chains, data flows, and type relationships across modules in a single pass.

## What to consider

GPT 5.3 Codex incorporates GPT-5.3 generation advances in reasoning and code comprehension over GPT-5.2 codex.

Each codex generation brings measurable improvements in code quality, reasoning depth, and task completion reliability.

## When to use

### Best For

- **Complex coding automation:** Tasks that benefit from GPT-5.3 generation reasoning improvements
- **Complex enterprise projects:** Large-scale development, migration, and modernization requiring strong model quality
- **Advanced agent architectures:** AI coding tools that benefit from GPT-5.3 generation improvements
- **Performance-critical code:** Optimization tasks where stronger reasoning produces meaningfully better results
- **Development tools:** Products that benefit from GPT-5.3 generation coding capabilities

### Consider Alternatives When

- **Routine coding tasks:** Earlier codex mini models for everyday bug fixes and scaffolding
- **Cost optimization:** Previous generation codex models at lower price points
- **Non-coding tasks:** GPT-5.2 or other general-purpose models for reasoning and generation workloads unrelated to code
- **Extended reasoning:** o3 for pure chain-of-thought reasoning on hard problems

## Best for

- **Complex coding automation:** Tasks that benefit from GPT-5.3 generation reasoning improvements
- **Complex enterprise projects:** Large-scale development, migration, and modernization requiring strong model quality
- **Advanced agent architectures:** AI coding tools that benefit from GPT-5.3 generation improvements
- **Performance-critical code:** Optimization tasks where stronger reasoning produces meaningfully better results
- **Development tools:** Products that benefit from GPT-5.3 generation coding capabilities

## Consider alternatives

- **Routine coding tasks:** Earlier codex mini models for everyday bug fixes and scaffolding
- **Cost optimization:** Previous generation codex models at lower price points
- **Non-coding tasks:** GPT-5.2 or other general-purpose models for reasoning and generation workloads unrelated to code
- **Extended reasoning:** o3 for pure chain-of-thought reasoning on hard problems

## Frequently asked questions

### How does GPT 5.3 Codex improve over GPT-5.2 codex?

It incorporates the GPT-5.3 generation's advances in reasoning and code comprehension, producing higher-quality code and more reliable autonomous task completion.

### What context window does GPT 5.3 Codex support?

400K tokens, enabling comprehensive codebase analysis in a single pass.

### Can GPT 5.3 Codex verify its own output?

Yes. It operates in sandboxed environments where it executes code, runs test suites, and iterates on solutions before returning results.

### How does AI Gateway handle authentication for GPT 5.3 Codex?

AI Gateway accepts a single API key or OIDC token for all requests. You don't embed OpenAI credentials in your application; AI Gateway routes and authenticates on your behalf.

### Should I upgrade from GPT-5.2 codex?

If your workloads benefit from improved reasoning and code quality, yes. Test with your specific tasks to confirm the improvement justifies any pricing differences.

### What are typical latency characteristics?

This page shows live throughput and time-to-first-token metrics measured across real AI Gateway traffic.

## Links

- [Model page](https://vercel.com/ai-gateway/models/gpt-5.3-codex)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://openai.com/index/introducing-gpt-5-3-codex/)
- [Provider pricing](https://openai.com/api/pricing/)