# GPT 5.4 Pro

The premium tier of the GPT-5.4 family, providing maximum performance for the most complex tasks that demand extended reasoning and thoroughness.

- **Model ID:** `openai/gpt-5.4-pro`
- **Type:** chat
- **Providers:** openai, azure
- **Context window:** 1,050,000
- **Maximum output tokens:** 128,000
- **Pricing:** $30/1M input tokens, $180/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/gpt-5.4-pro

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

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

const result = streamText({
  model: 'openai/gpt-5.4-pro',
  prompt: 'Why is the sky blue?'
})
```

## About

GPT 5.4 Pro became available on March 5, 2026 on AI Gateway as the premium tier of the GPT-5.4 model family. It's designed for developers who need maximum performance on the most complex tasks, applying extended compute to produce the deepest reasoning and highest-quality output.

Like the base GPT-5.4, it brings the agentic and reasoning leaps from GPT-5.3 Codex to all domains. The pro tier goes further, handling the subset of problems where the standard model's capability ceiling is a binding constraint. Complex multi-step workflows, advanced research synthesis, and high-stakes analysis all benefit from the additional reasoning depth.

With a context window of 1.1M tokens and the full API feature set, GPT 5.4 Pro handles any task the GPT-5.4 family supports, with more computation applied per request. Route your hardest queries here and use the standard tier for routine traffic.

## What to consider

GPT 5.4 Pro is for developers who need maximum performance on the most complex tasks. It applies the most compute per request in the GPT-5.4 family.

Most teams route only their hardest queries to the pro tier, pairing it with GPT-5.4 or GPT-5.4 mini for routine traffic to manage costs.

## When to use

### Best For

- **Mission-critical analysis:** High-stakes decisions that require the deepest available reasoning
- **Complex agentic workflows:** Multi-step tasks with tools, research, and multiple sources that need maximum reliability
- **Expert-level content creation:** Technical documentation, legal drafting, and precision-critical writing
- **Advanced research synthesis:** Comprehensive analysis of complex topics requiring exhaustive deliberation
- **Quality-gated pipelines:** Final-pass review where maximum quality justifies premium cost

### Consider Alternatives When

- **Standard workloads:** GPT-5.4 for most tasks without the pro tier's premium
- **Cost-sensitive applications:** GPT-5.4 mini or nano for high-volume production
- **Specialized reasoning:** o3-pro for pure mathematical and scientific chain-of-thought
- **Speed-critical responses:** GPT-5.4 mini or nano when latency is the priority

## Best for

- **Complex agentic workflows:** Multi-step tasks with tools, research, and multiple sources
- **Expert-level content creation:** Technical documentation, legal drafting, and precision-critical writing
- **Maximum GPT-5.4 compute:** Applies the most compute per request of any GPT-5.4 variant
- **Inherits GPT-5.3 Codex leaps:** Brings the agentic and reasoning gains from GPT-5.3 Codex into the GPT-5.4 pro tier

## Consider alternatives

- **Standard workloads:** GPT-5.4 for most tasks without the pro tier's premium
- **Cost-sensitive applications:** GPT-5.4 mini or nano for high-volume production
- **Specialized reasoning:** o3-pro for pure mathematical and scientific chain-of-thought
- **Speed-critical responses:** GPT-5.4 mini or nano for lower-latency workloads

## Frequently asked questions

### How does GPT 5.4 Pro differ from standard GPT-5.4?

It applies more compute per request, producing deeper reasoning and higher-quality output on hard problems. On simpler tasks the difference may be minimal.

### When is the pro tier justified?

For high-stakes analysis, complex multi-step agentic workflows, expert-level content, and tasks where the quality difference over standard GPT-5.4 directly impacts outcomes.

### What context window does GPT 5.4 Pro support?

1.1M tokens, matching the GPT-5.4 family.

### Is GPT 5.4 Pro slower than standard GPT-5.4?

It may take longer per request due to the additional compute. For latency-sensitive applications, route only specific high-value queries to the pro tier.

### Does GPT 5.4 Pro share the same agentic improvements as GPT-5.4?

Yes. It includes all of GPT-5.4's agentic and reasoning advances, plus extended compute for maximum performance on the hardest problems.

### How does AI Gateway handle authentication for GPT 5.4 Pro?

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.

### 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-5.4-pro)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://platform.openai.com/docs/models/gpt-5.4-pro)
- [Provider pricing](https://openai.com/api/pricing/)


If you encounter incorrect information, can’t find what you need, need a capability Vercel doesn’t currently offer, or have any other feedback to report, please do it by following the instructions at https://vercel.com/feedback.md.