A support team auto-tagging incoming tickets
- Create an API key in the OpenAI platform
- Send each ticket text to gpt-6-luna with a list of allowed tags
- Store the returned tag in the help-desk system
needs API key
OpenAI · API platform
OpenAI API
Pay-per-token developer access to OpenAI's GPT models for building apps, agents and automations.
- Provider
- OpenAI
- Latest confirmed model
- GPT-6 Astra (gpt-6-astra)
- Pricing
- Paid only
- Free plan
- No
- API
- Yes
- Status
- Generally available
- Last verified
- Oct 19, 2026
What is OpenAI API?
The OpenAI API lets developers send text, images and other inputs to OpenAI's GPT models from their own code and receive generated responses. You create an API key, call the models over HTTPS or with an official SDK, and pay only for the tokens you use. The current lineup is split into a most-capable model (GPT-6 Astra), a mid-tier model (GPT-6.1 Sol) and a low-cost high-volume model (GPT-6 Luna).
Adding language understanding, generation and reasoning to a product without training or hosting a model yourself.
Why use OpenAI API?
Practical benefits
- Several price/performance tiers from the same provider
- Large 1.05M-token context window on current GPT-6 models
- Batch and Flex processing at roughly half the standard rate
- Cached input billed at a fraction of normal input price
Who it suits
- Software developers
- Startups building AI features
- Teams automating document or support workflows
When another tool may fit better
- People who want a ready-made chat app with no coding (use ChatGPT)
- Workloads that must run fully offline or on your own hardware
- Users who need a guaranteed free API allowance
- Learning curve
- Some experience helps
- Skills needed
- Basic programming and HTTP/JSON; understanding of tokens and prompt design helps.
Example workflows
- Generate an API key, install the SDK, call gpt-6-luna to classify support tickets
- Send large overnight jobs through the Batch API to cut cost
What can it do?
- Tiered GPT-6 model familyAstra (most capable), Sol (near-Astra at lower cost) and Luna (efficient, high-volume).
- Batch processingAsynchronous jobs billed at roughly 50% of standard rates.
- Flex processingLower-priority processing priced at batch rates for supported models.
- Fast modeHigher-priority processing (formerly Priority) at roughly 2x standard price for many models.
- Prompt cachingRepeated input billed at a discounted cached-input rate.
- Moderation modelomni-moderation-latest is listed as free to use.
Good for
- Chatbots and assistants inside products
- Document summarisation and extraction
- Coding and agent workflows
- Content classification and moderation
Where it runs
Pricing and plans
Paid only Usage-based billing per 1M tokens, with separate rates for input, cached input and output. GPT-6 Astra costs $10 input / $50 output, GPT-6.1 Sol $2 / $10 and GPT-6 Luna $0.10 / $0.50 per 1M tokens on the Standard tier. Batch and Flex are about half price; Fast mode costs more.
| Plan | Price | Billing | Free trial / tier | Usage limits | Main inclusions | Model access | API access | Restrictions |
|---|---|---|---|---|---|---|---|---|
| Standard (pay as you go) | Billed per 1M tokens by model; prices shown are for inputs up to 272K context. | usage-based | — | Rate limits depend on account usage tier. |
|
Full model lineup including gpt-6-astra, gpt-6.1-sol, gpt-6-luna and older GPT-5.x/4.x models | Yes | Prepaid or billed usage required |
| Batch | Roughly 50% of Standard rates (e.g. gpt-6-astra input $5 per 1M). | usage-based | — | Asynchronous; results are not immediate. |
|
Listed models | Yes | Not suitable for real-time responses |
| Flex | Same rates as Batch for models in the Flex table. | usage-based | — | Lower-priority processing. |
|
GPT-5.x and some older models per the pricing table | Yes | Not all models supported |
| Fast mode | Roughly 2x Standard for many models (e.g. gpt-6-astra input $20 per 1M). | usage-based | — | — |
|
Listed models | Yes | — |
Worked example (usage-based API pricing): sending 1,000,000 input tokens and receiving 200,000 output tokens at the listed gpt-6-astra rates ($10 and $50 per 1M) costs about $20 before tax. Real bills depend on the model, caching, batch discounts and any tier you qualify for.
API and developer access
- Official API
- Yes
- Documentation
- API docs (opens in a new tab)
- Authentication
- API key (Bearer)
- Pricing model
- per 1M tokens
- Free API quota
- None documented — the free plan above (if any) is for the app, not the API
- Rate limits
- Tier-based; varies by account
- Context / input limits
- 1.05M tokens (gpt-6-astra, gpt-6.1-sol, gpt-6-luna); 128K max output
- SDKs
- Not documented
- Streaming
- Not documented
- Batch processing
- Yes
- Structured output
- Not documented
- Tool / function calling
- Not documented
- Commercial use
- See the provider terms
- Data retention & privacy
- See the provider privacy terms
Published API rates
| Item | Price | Unit |
|---|---|---|
| gpt-6-astra input | $10 | per 1M input tokens |
| gpt-6-astra cached input | $1 | per 1M input tokens |
| gpt-6-astra output | $50 | per 1M output tokens |
| gpt-6.1-sol input | $2 | per 1M input tokens |
| gpt-6.1-sol output | $10 | per 1M output tokens |
| gpt-6-luna input | $0.10 | per 1M input tokens |
| gpt-6-luna cached input | $0.01 | per 1M input tokens |
| gpt-6-luna output | $0.50 | per 1M output tokens |
A free OpenAI API account is not the same thing as free API access. Unless a free API quota is listed above, API calls are billed separately.
Advantages and disadvantages
Advantages
- Clear model tiers with published per-token prices
- Very large context window on current models
- Multiple cost-saving modes (Batch, Flex, caching)
Disadvantages
- No documented free API tier for the GPT models
- Costs can grow quickly at the top tier ($50 per 1M output tokens for Astra)
- Higher pricing applies above 272K input tokens per the pricing page tables
Privacy: Check OpenAI's API data-usage and retention policies for your account; not verified in this research.
Real-world examples
A developer summarising thousands of reports overnight
- Prepare a JSONL file of requests
- Submit it through the Batch API
- Download results when the batch completes, paying about half the standard rate
needs API key
A startup building a research assistant
- Use gpt-6.1-sol for most queries
- Route the hardest questions to gpt-6-astra
- Cache the long shared system prompt to reduce input cost
needs API key
OpenAI API alternatives
Claude Developer Platform (Claude API)
Developer access to Anthropic's Claude models with usage-based pricing, batch discounts and prompt caching.
OpenAI API vs Claude Developer Platform (Claude API)Gemini API and Google AI Studio
Google's developer API for Gemini models, with a free tier in Google AI Studio and paid tiers for production.
OpenAI API vs Gemini API and Google AI StudioOpenRouter
One OpenAI-compatible API key for hundreds of models from many providers.
OpenAI API vs OpenRouterTogether AI
Cloud platform for running, fine-tuning and training open models on serverless APIs or rented GPUs.
OpenAI API vs Together AIGroqCloud
Low-latency inference API for open models running on Groq's LPU hardware.
OpenAI API vs GroqCloudFrequently asked questions
Is the OpenAI API free?
No free API tier for the GPT models is described on the official pricing page. You pay per token; only the omni-moderation-latest model is listed as free.
Which OpenAI model is the most capable?
OpenAI's models page describes GPT-6 Astra (gpt-6-astra) as its most capable model, priced at $10 per 1M input and $50 per 1M output tokens.
What is the cheapest current model?
GPT-6 Luna (gpt-6-luna) is described as the most efficient model, at $0.10 per 1M input and $0.50 per 1M output tokens. Older gpt-5-nano is also listed at $0.05 / $0.40.
How can I cut costs?
Use the Batch API or Flex processing (about half the standard rate), reuse prompts to benefit from cached-input pricing, and choose a smaller model for simple tasks.
How large is the context window?
The current GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna models list a 1.05M-token context window and 128K max output.
Is the API the same as ChatGPT?
No. ChatGPT is a consumer app with its own subscriptions; the API is billed separately per token and is meant for developers.