What for: direct access to GPT and OpenAI's multimodal APIs — text, vision, images, voice, embeddings — straight from the source.

What it is and who it's for

OpenAI Platform is direct access to the GPT models and a broad set of multimodal APIs: text, vision, image generation (GPT Image), voice (TTS/Realtime), embeddings, and Assistants/Responses for agents. It's for developers who want one ecosystem that covers everything, with rich tooling and mature SDKs. GPT models sit consistently in the top group of the blind LMArena, and the platform stands out for offering the widest range of modalities and features from a single account.

Key features

The GPT family (including mini variants for cheap, fast tasks) behind one API.
Multimodal: vision, GPT Image (images), TTS and the Realtime API (voice), Whisper (STT).
Function calling, structured output (JSON Schema), Responses/Assistants for agents.
Embeddings for search and RAG; the Batch API for cheap bulk processing.
Prompt caching on repeated context; official SDKs (Python/TS).
A huge ecosystem of examples, tools and integrations.

Get started in 5 minutes

Create a key at platform.openai.com.
Call a model:

curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.2","messages":[{"role":"user","content":"Hi"}]}'

Need images or voice? Hit the matching endpoint (Images / Audio) with the same key.

When to use it, when not to

✅ Use it if you want one ecosystem covering text + images + voice + embeddings with rich tooling.
✅ Use it if you're building agents or RAG and value mature SDKs and examples.
❌ Skip it → the Anthropic API is better if top-tier code and reasoning quality plus long context are the priority.
❌ Skip it → OpenRouter is better if you want to switch cheaply between providers on one key and have fallbacks.

Honest pricing

Pay per token/request, with rates varying by model and modality; mini models plus Batch and caching cut the bill noticeably. Exact numbers are on the provider's site and keep changing.

Gotchas

Top models cost more — use the mini variants for simple work instead of overpaying.
Different modalities are billed differently — budget for your specific scenario.
Names, versions and deprecations keep moving — pin the version in your config.
It's a cloud: sensitive data leaves your perimeter — check your compliance requirements.

🤖 Prompt accelerator

"I'm building an <app/agent> on OpenAI Platform with text and possibly images/voice. Help me pick a model for the task and my budget (top-tier vs mini), show Chat Completions calls with function calling and structured output in <Python/Node>, and demonstrate how to plug in the Images or Realtime API with the same key. Estimate the cost of a typical request and where to switch on Batch or caching."

OpenAI Platform (GPT)