OpenAI

OpenAI is the San Francisco lab behind ChatGPT and the GPT API. This page used to stop at GPT-4, o1, DALL-E 3, and Sora as if they were the current catalog. GPT-5.6 Sol is the frontier recommendation. GPT-5.2 is still on the API as the previous frontier snapshot.

The company still sells research, a consumer chat product, and a developer platform. Treat those as three surfaces, not one model.

Key Features

• GPT-5.6 family: Sol for frontier work, plus Terra and Luna on the same generation. Sol docs list a 1,050,000-token context.
• Previous snapshot still billed: GPT-5.2 at $1.75 input / $14 output per 1M tokens, 400k context, 128k max out.
• Realtime and voice: Agents SDK docs now call gpt-realtime-2.1 for realtime agents.
• Developer surface: platform.openai.com for keys; developers.openai.com for model cards and price.
• Agent runtime: official OpenAI Agents SDK in Python and TypeScript.

Use Cases

• New API builds that should start on GPT-5.6 Sol unless a pin requires GPT-5.2.
• Consumer chat via ChatGPT rather than the raw API.
• Agent products that want OpenAI's own runtime instead of a third-party graph.

Limitation: ChatGPT Plus marketing pages often 403 from scrapers. DALL-E 3 is no longer the image story; see GPT Image 2 and the DALL-E 3 deprecation note.

Pricing

Verified API list prices on 2026-08-17. Confirm the pricing page before you quote them in a contract.

Getting Started

1. Create a key on platform.openai.com.
2. Read the models index and pick GPT-5.6 Sol unless you must pin GPT-5.2.
3. For agents, install the Agents SDK instead of wrapping chat completions by hand.
4. For consumer use, go to ChatGPT, not the API dashboard.

Frequently Asked Questions

Is GPT-4 still the flagship?

No. Docs recommend GPT-5.6 Sol. GPT-5.2 is the previous frontier.

Is DALL-E 3 the current image model?

No. This catalog already marks DALL-E 3 as deprecated. Use the current image entry.

Where do I read live prices?

developers.openai.com/api/docs/pricing, not launch blogs.

Alternatives

• Anthropic: Claude line if you want a non-OpenAI frontier.
• Kimi: Moonshot consumer + API if you want a 1M-context non-US stack.
• Zhipu GLM-5.3: Open-weight coding alternative.

Tips

1. Quote Sol and 5.2 from the model pages, not from this company overview.
2. Keep consumer ChatGPT and API billing in different sentences.
3. Confirm the models index. Family names moved after GPT-4 / o1.

Conclusion

OpenAI is still the default US lab for chat and API work, but the live catalog is GPT-5.6 Sol plus a billed GPT-5.2 snapshot, not GPT-4 and DALL-E 3. Start at the models index, then open GPT-5.6 Sol or GPT-5.2.