The OpenAI API (also called the GPT API) is OpenAI's developer platform — documented at developers.openai.com and formerly at platform.openai.com — for calling GPT and reasoning models programmatically instead of through the ChatGPT interface. It's built for developers and businesses embedding AI into their own products, and it bills strictly pay-as-you-go: you pay per input and output token, at rates that vary by model rather than a flat monthly fee.
Who builds it
The OpenAI API is built and operated by OpenAI. It runs on the same underlying models as ChatGPT, but is a distinct product with its own API keys, dashboard, and billing, aimed at developers rather than end users chatting directly.
Core features
- Pay-as-you-go token billing — every request is metered by input and output tokens and billed at the selected model's per-million-token rate, with no flat subscription fee.
- A model lineup for every budget — OpenAI's current flagship family, GPT-5.6, ships in three tiers: Luna ($1/$6 per million input/output tokens) for cost-sensitive volume work, Terra ($2.50/$15) balancing cost and capability, and Sol ($5/$30), the frontier model for complex professional and coding tasks.
- Dedicated reasoning models — the o-series (o3, o4-mini, o3-pro) stays available through the API for multi-step reasoning, math, and coding, even after being retired from the ChatGPT interface in February 2026.
- Responses API — OpenAI recommends its newer Responses API over the older Chat Completions API for building new text-generation applications.
- Automatic usage-tier scaling — rate limits (requests and tokens per minute) increase automatically as an account's cumulative spend rises, from a $100/month cap at the entry tier up to $200,000/month at the top tier.
- API-key based access — requests are authenticated with project or organization API keys, with separate admin keys and workload identity tokens for enterprise setups, entirely independent of a ChatGPT login.
Pricing
The OpenAI API is billed per model, per token — there is no single flat price. As of OpenAI's current pricing page, GPT-5.6 Luna costs $1 per million input tokens and $6 per million output tokens, GPT-5.6 Terra costs $2.50/$15, and GPT-5.6 Sol costs $5/$30. The o3 reasoning model is priced separately at $2 per million input tokens ($0.50 for cached input) and $8 per million output tokens. New accounts are capped at roughly $100 of monthly usage until their spend qualifies them for a higher tier; there is no ongoing free quota for production use.
Who it's for
The OpenAI API is for developers and companies building their own applications, chatbots, agents, or automated workflows on top of GPT models — not for people who simply want to chat, who should use ChatGPT instead. It suits teams that need programmatic control over model choice, prompts, and cost, and that are comfortable managing API keys, rate limits, and usage-based billing themselves. Because pricing and capability are split across many model tiers, it also suits teams that want to fine-tune the cost-versus-quality tradeoff per feature — for example, routing simple classification calls to a cheaper model like GPT-5.6 Luna while reserving GPT-5.6 Sol or an o-series reasoning model for harder, higher-value requests.