OpenAI Guide
OpenAI is no longer "just ChatGPT". It is a consumer and business assistant, a full developer platform, a coding agent stack, and a team workspace with governance features. This guide maps the surfaces, shows the shortest path to a first API call, and helps you choose the right OpenAI entry point.
Based on official OpenAI sources across chatgpt.com, openai.com, and developers.openai.com. The current flagship model across the public docs is GPT-5.5, the preferred programmatic path is the Responses API, and ChatGPT now includes an agent mode that can act on the web. Product plans, limits, and model availability change frequently, so re-check live plan and model pages before quoting exact entitlements.
1. The mental model
| Surface | What it is for | Primary user |
|---|---|---|
| ChatGPT | Daily AI assistant for writing, analysis, web search, images, files, voice, and agentic tasks | Individuals, knowledge workers |
| ChatGPT Business / Enterprise / Edu | Shared ChatGPT workspace with admin controls, governance, and business privacy commitments | Teams, IT, compliance-driven orgs |
| OpenAI API | Programmatic access to models, tools, multimodal input/output, and production app building | Developers, product teams |
| Codex | OpenAI's coding assistant across app, editor, and terminal surfaces | Engineers, technical teams |
| Apps / Workspace Agents | Extend ChatGPT with tools, actions, apps, and published workspace agents | Internal platform teams, advanced builders |
Rule of thumb:
- Need an AI assistant for daily work? Start with ChatGPT.
- Need to build an application or automation? Start with the OpenAI API.
- Need repo-aware coding help? Use Codex.
- Need admin controls and business privacy? Use ChatGPT Business or Enterprise.
2. ChatGPT
ChatGPT is the fastest way to get value from OpenAI if you are not writing code yet.
Typical use cases:
- drafting and editing text,
- asking search-grounded questions with web results,
- analyzing uploaded files,
- generating images,
- speaking through voice mode,
- handing off multi-step tasks to the agent feature.
The current public ChatGPT product surface also includes:
- Apps for integrating external tools,
- Canvas for collaborative editing,
- Images and multimodal conversations,
- Codex access on supported plans,
- Agent for tasks that can browse and act on the web.
OpenAI explicitly frames ChatGPT agent as a higher-risk feature because it can act on websites and work with connected data. Treat it as a supervised assistant, not as an unsupervised operator.
3. The OpenAI API
The API is the build-on surface. It is where you create applications, automations, internal tools, and product features.
The current API stack revolves around the Responses API, which OpenAI positions as the preferred path for modern model interactions.
What the API supports out of the box:
- text generation,
- vision and image input,
- audio and speech workflows,
- structured outputs,
- function calling,
- built-in tools like web search and code interpreter,
- MCP and remote connectors,
- stateful, tool-using, agentic workflows.
OpenAI's own docs currently recommend:
- GPT-5.5 for complex reasoning and coding,
- GPT-5.4 mini or GPT-5.4 nano for lower-latency and lower-cost workloads.
If you are still on Chat Completions for new work, the docs make it clear that Responses is the more future-facing interface.
4. Codex
Codex is OpenAI's dedicated coding surface. It is no longer only a model name from the past; it is now a product and workflow layer for engineering work.
The official positioning is straightforward:
- Codex for work: gets tasks and projects done from start to finish,
- Codex for developers: built for real engineering work across multiple surfaces.
That matters because Codex sits between a normal chat assistant and a full developer platform:
- it works with real code and project context,
- it spans app, editor, and terminal usage,
- it is meant for repeatable engineering workflows, not just one-off snippets.
Choose Codex when the job is not "answer my question", but "read the repo, make the change, and help me ship it."
5. Business, governance, and privacy
For team usage, the main split is:
| Need | Recommended surface |
|---|---|
| personal productivity | ChatGPT |
| team collaboration with admin controls | ChatGPT Business |
| larger org governance and enterprise rollout | ChatGPT Enterprise |
| product integration and backend automation | OpenAI API |
OpenAI's enterprise privacy page currently states two especially important commitments for business products and the API platform:
- data ownership and control remain with the customer,
- model training on business data is not enabled by default.
That does not mean "send anything anywhere without review". It does mean OpenAI has a clearer enterprise operating story than a consumer-only chatbot vendor.
6. Quickstart (zero -> first API call)
- Create an API key in the OpenAI dashboard.
- Export it as an environment variable:
export OPENAI_API_KEY="your_api_key_here"
- Send a first request with the Responses API:
from openai import OpenAI
client = OpenAI()
response = client.responses.create(
model='gpt-5.5',
input='Write a one-sentence bedtime story about a unicorn.',
)
print(response.output_text)
This is the current "golden path" in the official docs for new API integrations.
7. Decision guide
| If you want to… | Use… |
|---|---|
| chat, write, summarize, search, and work with files | ChatGPT |
| give a team a shared, governed AI workspace | ChatGPT Business or Enterprise |
| build a product or backend workflow | OpenAI API |
| build tool-using or agentic workflows | Responses API and the Agents stack |
| work directly on repos, diffs, tests, and engineering tasks | Codex |
| extend ChatGPT with tools or internal integrations | Apps or Workspace Agents |
8. Recommended starting points
- Individual knowledge worker: start with ChatGPT.
- Developer building a product: start with the OpenAI API and the Responses API.
- Engineering team: pair ChatGPT or Business with Codex for implementation work.
- IT / platform / compliance team: evaluate ChatGPT Business or Enterprise first, then define which API use cases are allowed.
9. Official links
Products
Developer
Related guides