---
title: "gpt-5.4 pricing and API | Inference.net"
description: "gpt-5.4 pricing, context limits, capabilities, and an OpenAI-compatible quickstart on Inference.net."
url: "https://inference.net/models/gpt-5.4/"
---

# GPT-5.4

OpenAI's multimodal model, served through one OpenAI-compatible endpoint with transparent per-token pricing and project-scoped API keys.

OpenAI Multimodal Model

[Get started](https://inference.net/register/) [Documentation](https://inference.net/docs/api/quickstart)

MODEL API PRICING

PER 1M TOKENS

Input

$2.50

Cache read

$0.25

Output

$15.00

[See all pricing](https://inference.net/subscription/)

MODEL SPEC

Developed by

OpenAI

Model family

GPT-5.4

Size

—

Architecture

Transformer

Context window

1.1M tokens

Max output

128K tokens

License

Provider license

Overview

GPT-5.4 is a multimodal model from OpenAI, available through the Inference.net serverless API. Use the same OpenAI-compatible request shape across the catalog and switch models by changing a single identifier.

The API supports streaming responses and project-scoped credentials. Usage, token volume, and spend appear in the dashboard automatically, so the model can move from evaluation to production without a second integration path.

QUICKSTART [Get API key](https://inference.net/dashboard/api-keys/)

```typescript
import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.inference.net/v1',
  apiKey: process.env.INFERENCE_API_KEY,
})

const completion = await client.chat.completions.create({
  model: 'gpt-5.4',
  messages: [{ role: 'user', content: 'Hello' }],
})
```

PROMPT CACHING

1. Keep the prompt prefix byte-identical across calls: same text, same order.
2. Pass a stable prompt_cache_key string. It is a routing-affinity hint; it does not create or identify cached content.
3. Check usage.prompt_tokens_details.cached_tokens in the response to confirm cache reads.

```
{
  "prompt_cache_key": "v1-stable-prefix",
  "messages": [{ "role": "user", "content": "..." }]
}
```

AGENT INTEGRATIONS [View all](https://inference.net/docs/integrations/overview/)

[### Claude Code Native Anthropic /v1/messages. Six model slots for every Claude Code workload. `fast claude on`](https://inference.net/docs/integrations/claude-code/) [### Codex OpenAI-compatible Responses API with a reversible Codex configuration. `fast codex on`](https://inference.net/docs/integrations/codex/) [### Grok Chat completions with a catalog-backed section for each supported alias. `fast grok on`](https://inference.net/docs/integrations/grok/) [### OpenCode Chat completions with Inference.net models registered as an OpenCode provider. `fast opencode on`](https://inference.net/docs/integrations/opencode/) [### Pi Chat completions with model and provider settings written for Pi Agent. `fast pi on`](https://inference.net/docs/integrations/pi/) [### Hermes Hermes provider configuration is documented while its CLI adapter is in preview. `CLI adapter in preview`](https://inference.net/docs/integrations/hermes/) [### OpenClaw OpenClaw provider routing is documented while its CLI adapter is in preview. `CLI adapter in preview`](https://inference.net/docs/integrations/openclaw/) [### Fx Fx gateway configuration is documented while its CLI adapter is in preview. `CLI adapter in preview`](https://inference.net/docs/integrations/fx/) [### Cursor Cursor base URL routing is documented while its CLI adapter is in preview. `CLI adapter in preview`](https://inference.net/docs/integrations/cursor/)

Ready when you are

## Start building with GPT-5.4 today

Serverless, OpenAI-compatible, and billed per token. No commitments.

[Get started](https://inference.net/register/) [Read the docs](https://inference.net/docs/api/quickstart)
