OpenAI's API powers the new flagship GPT-6 Astra (launched September 3, 2026), the GPT-5.6 family (Sol/Terra/Luna), GPT-4o and the o-series reasoning models — the most widely used large-language-model API in production. Here's what it costs, what you get for free, and exactly how to get your key.
| Model | Input $/1M | Output $/1M | Best for |
|---|---|---|---|
| GPT-6 Astra flagship | $10.00 | $50.00 | Deepest reasoning, highest cost |
| GPT-6.1 Sol new | $2.00 | $10.00 | Cheaper than GPT-5.6 Sol's own discount rate |
| GPT-6 Luna new | $0.10 | $0.50 | Default for routing/classification/extraction |
| GPT-5.6 Sol promo | $4.00 | $20.00 | Temporary discount rate, through at least Nov 21, 2026 |
| GPT-5.6 Terra | $2.00 | $12.00 | Flagship reasoning + generation |
| GPT-5.6 Luna | $0.20 | $1.20 | Superseded by GPT-6 Luna above |
| GPT-5 nano cheapest legacy | $0.05 | $0.40 | Highest volume, simple tasks |
| GPT-4o | $2.50 | $10.00 | Vision, still widely used |
| GPT-4o mini | $0.15 | $0.60 | Cheap legacy workhorse |
| o3 | $2.00 | $8.00 | Hard reasoning (repriced) |
→ Estimate your bill with the GPT-6 Astra API cost calculator (with a GPT-5.6 Sol savings comparison), the GPT-4o API cost calculator, or compare every model on the full AI API cost calculator.
OpenAI has no permanent free tier for the API. New accounts occasionally get a small amount of trial credit that expires, but ongoing usage is pay-as-you-go. The good news: the cheapest model, GPT-5 nano, is so inexpensive that light projects often cost a few cents per day. Want a real free tier instead? See Google Gemini, which offers a generous free quota on AI Studio.
1. Go to platform.openai.com and create an account (or sign in).
2. Open Settings → Billing and add a payment method. The API won't work without billing set up.
3. Go to the API keys page (platform.openai.com/api-keys).
4. Click Create new secret key, give it a name, and copy it — you can only see it once.
5. Set a monthly usage limit under Billing → Limits so a runaway script can't surprise you.
Test it with a simple request:
Use the AI API Cost Calculator to plug in your token counts and request volume — it shows your monthly cost and ranks every model (including Claude and Gemini) from cheapest to most expensive for your workload.
If GPT-6 Astra is more than you need — and for most everyday chat, summarization or extraction it is — drop to GPT-5.6 Sol (2.5× cheaper) or all the way to GPT-5.6 Luna / GPT-5 nano (≈15–100× cheaper) for most tasks; or compare Claude Haiku and Gemini Flash, which target the same low-cost, high-volume niche. For the absolute cheapest, DeepSeek-V3 undercuts most US providers.
No permanent free tier. New accounts may get small trial credits that expire; ongoing use is pay-as-you-go. GPT-5 nano is the cheapest model at ~$0.05/1M input tokens.
Create an account at platform.openai.com, add billing, open the API keys page, click Create new secret key, copy it once, and set a monthly usage limit.
GPT-5 nano — $0.05/1M input, $0.40/1M output — roughly 3× cheaper than GPT-4o mini and 100× cheaper than the GPT-5.6 Sol flagship, and good enough for most classification, extraction and simple generation tasks.
Generating tokens is more compute-intensive than reading them, so output is billed at 3–5× the input rate across providers.
Not affiliated with OpenAI. Prices are reference estimates — always verify on the official pricing page.