18 models tracked
OpenAI API pricing
OpenAI publishes its tokenizers, which makes it the only provider here where a token count taken in your browser is exact rather than estimated.
- Models tracked
- 18 11 current
- Cheapest input
- $0.0500 GPT-5 nano
- Highest input
- $5.00 GPT-5.5
- Exact token counts
- 18 / 18 tokenizer published
Every OpenAI model, cheapest first
| Model | Input | Cached | Output | Context | Count | Chat turn |
|---|---|---|---|---|---|---|
| GPT-5 nano | $0.0500 | $0.005000 | $0.4000 | — | Exact | $0.000195 |
| GPT-4o mini legacy | $0.1500 | $0.0750 | $0.6000 | — | Exact | $0.000405 |
| GPT-5.6 Luna | $0.2000 | $0.0200 | $1.20 | 1.05M | Exact | $0.000660 |
| GPT-5.4 nano | $0.2000 | $0.0200 | $1.25 | — | Exact | $0.000675 |
| GPT-5 mini | $0.2500 | $0.0250 | $2.00 | — | Exact | $0.000975 |
| GPT-4.1 mini legacy | $0.4000 | $0.1000 | $1.60 | — | Exact | $0.001080 |
| GPT-3.5 Turbo legacy | $0.5000 | — | $1.50 | — | Exact | $0.001200 |
| GPT-5.4 mini | $0.7500 | $0.0750 | $4.50 | — | Exact | $0.002475 |
| o4-mini legacy | $1.10 | $0.2750 | $4.40 | — | Exact | $0.002970 |
| GPT-5.1 | $1.25 | $0.1250 | $10.00 | — | Exact | $0.004875 |
| GPT-5 | $1.25 | $0.1250 | $10.00 | — | Exact | $0.004875 |
| GPT-4.1 legacy | $2.00 | $0.5000 | $8.00 | — | Exact | $0.005400 |
| o3 legacy | $2.00 | $0.5000 | $8.00 | — | Exact | $0.005400 |
| GPT-4o legacy | $2.50 | $1.25 | $10.00 | — | Exact | $0.006750 |
| GPT-5.6 Terra | $2.00 | $0.2000 | $12.00 | 1.05M | Exact | $0.006600 |
| GPT-5.4 | $2.50 | $0.2500 | $15.00 | — | Exact | $0.008250 |
| GPT-5.6 Sol | $5.00 | $0.5000 | $30.00 | 1.05M | Exact | $0.0165 |
| GPT-5.5 | $5.00 | $0.5000 | $30.00 | — | Exact | $0.0165 |
Chat turn is 1,500 input and 300 output tokens. Prices read from OpenAI’s pricing documentation; each row links to its own dated source.
What is distinctive about OpenAI pricing
- 01
Cached input is discounted heavily — commonly a tenth of the base rate on current models — which makes a stable system prompt the single largest saving available on a repetitive workload.
- 02
The model line spans a wide range: the gap between the cheapest and dearest text model is more than two orders of magnitude, so tier selection matters more here than provider selection.
- 03
Reasoning models bill their internal thinking as output tokens you never see, which makes budgets built on visible answer length wrong in the same direction every time.
Token counting on OpenAI
Two encodings are in use: o200k_base for GPT-4o and everything after it, and cl100k_base for GPT-4 and GPT-3.5. The newer one produces ten to twenty percent fewer tokens for the same text, and the gap is larger on code and non-English content.
The token counter runs the real encoder in your browser for these models, so the figure it gives is the one you will be billed for.
Compare with other providers
Or shortlist across all of them at once in the model finder, filtering by budget, context window and whether token counts can be exact.
Frequently asked questions
- How much does the OpenAI API cost?
- It depends on the model. Across the 18 models tracked here, input ranges from $0.0500 to $5.00 per million tokens and output from $0.4000 to $30.00. On a typical chat turn of 1,500 input and 300 output tokens the range is $0.000195 to $0.0165 per request.
- What is the cheapest OpenAI model?
- GPT-5 nano, at $0.0500 per million input tokens and $0.4000 per million output. Whether it is cheap enough depends on whether it can do your task — the cost of finding out is usually a few cents.
- Can I count OpenAI tokens exactly?
- Yes. OpenAI publishes its tokenizers, so counts taken before you send are exact for 18 of the 18 models tracked here.
- How current is this OpenAI pricing?
- Every model entry stores the URL it was read from and the date it was read, both shown in the table above. The full data set was last reviewed on August 5, 2026. Prices change without notice, so verify against OpenAI's own documentation before committing spend.