OpenAI pricing tiers
“OpenAI pricing tiers” means two different things, and mixing them up is the most common cause of a surprise bill. There are subscription tiers(Free, Plus, Pro…) for ChatGPT, and usage tiers (Tier 1 to Tier 5) that govern your API rate limits — not your per-token price. Below are both, plus the actual per-model API rates.
API usage tiers (Tier 1–5)
Usage tiers set how many requests per minute and tokens per minute you can send. The per-token price is the same on every tier — moving up a tier unlocks throughput, not a discount.
| Tier | Qualification | Rate limit |
|---|---|---|
| Free | $0 | 3 RPM |
| Tier 1 | $5 paid | 500 RPM |
| Tier 2 | $50 paid + 7 days | 5,000 RPM |
| Tier 3 | $100 paid + 7 days | 5,000 RPM |
| Tier 4 | $250 paid + 14 days | 10,000 RPM |
| Tier 5 | $1,000 paid + 30 days | 10,000 RPM |
Limits are per-model and are raised automatically as spend accumulates. Exact numbers change; check your account’s limits page before capacity planning.
ChatGPT subscription tiers
Flat monthly plans for using ChatGPT as a product. These do not include API credits — a Plus subscription gives you zero API calls, and API spend does not raise your ChatGPT limits.
| Plan | Price |
|---|---|
| Free | $0 / month |
| Go | ~$8–20 / month (regional) |
| Plus | $20 / month |
| Pro | $200 / month |
| Team | per-seat, billed annually/monthly |
| Enterprise | custom |
Plan names and prices vary by region and change often — treat these as orientation, not a quote.
API price per model (per 1M tokens)
Prices verified 2026-10-20. For any current model, input tokens are the cheap side and output tokens cost several times more — see what 1,000 words of output costs to put that in intuitive units.
| Model | Input / 1M | Cached input / 1M | Output / 1M | Context | Cost per 1k req* |
|---|---|---|---|---|---|
| GPT-4o | $2.50 | $1.25 | $10.00 | 128K | $7.50 |
| GPT-5 | $1.25 | $0.125 | $10.00 | 400K | $6.25 |
| GPT-5 mini | $0.250 | $0.025 | $2.00 | 400K | $1.25 |
| GPT-4o mini | $0.150 | $0.075 | $0.600 | 128K | $0.450 |
| GPT-5 nano | $0.050 | $0.0050 | $0.400 | 400K | $0.250 |
*1,000 requests at 1,000 input + 500 output tokens each — a typical short chat turn with some context.
Which tier do you actually need?
If you are building a product, the subscription tiers are irrelevant — ignore Plus and Pro entirely and budget on API rates. Start at Tier 1; the free tier is only for evaluation. The practical question is whether your monthly token volume fits the cheapest model that passes your quality bar, which is what the prompt cost estimator answers for a real workload shape.
If you are an individual deciding between a subscription and paying for API access: a $20 Plus plan is roughly equal to 10–20M tokens of a small model, or about 2–4M tokens of a frontier model. Below that usage, an API key plus a lightweight client is often cheaper; above it, the subscription usually wins unless you need the API specifically.
Frequently asked
- Does a higher usage tier lower my price?
- No. Per-million-token prices are identical across Tier 1–5. Tiers only raise rate limits. For cheaper tokens you switch model or use prompt caching.
- Do Plus or Pro subscriptions include API credit?
- No. ChatGPT plans and API billing are separate products on separate statements.
- How do cached inputs change the tier math?
- Cached input is billed at a fraction of normal input. If most of your prompt is a fixed system message or document, caching can cut total cost by more than switching to a cheaper model — the estimator has a cache-hit slider to model it.