OpenAI API cost calculator
Enter your request volume and token counts. Cached input tokens go in the cache read field; OpenAI has no cache write fee, so leave cache write at zero.
GPT-5.4 at 1,000 requests per day costs $0.0165 per request, $16.50 per day, $495.00 per month and $6,022.50 per year.
Prices effective 2026-09-06Verified 30 day month, 365 day year
| Line | Quantity per request | Unit price | Per request | Per month |
|---|---|---|---|---|
| input | 3,000 | $2.50 per 1m tokens | $0.007500 | $225.00 |
| output | 600 | $15.00 per 1m tokens | $0.009000 | $270.00 |
- Chat assistant (GPT-5.4 mini, 2k in / 400 out)
- Agent with 30k cached context (GPT-5.4)
- Bulk extraction (GPT-5.4 nano, batch)
Same workload on other models
| Model | Provider | Per request | Per month | Change from selected |
|---|---|---|---|---|
| GPT-5 nano | OpenAI | $0.000390 | $11.70 | -$483.30 |
| GPT-4.1 nano | OpenAI | $0.000540 | $16.20 | -$478.80 |
| GPT-4o mini | OpenAI | $0.000810 | $24.30 | -$470.70 |
| GPT-5.6 Luna | OpenAI | $0.001320 | $39.60 | -$455.40 |
| GPT-5.4 nano | OpenAI | $0.001350 | $40.50 | -$454.50 |
| GPT-5 mini | OpenAI | $0.001950 | $58.50 | -$436.50 |
Full ranking on the comparison page. Want CI to fail when a change pushes the monthly figure past a limit? See budget contracts.
How the number is computed
- Each token dimension is multiplied by its list price per 1M tokens in integer micro-dollars, then rounded half up once per line. No floating point.
- Cache write tokens use the 5 minute TTL price. Cache read tokens use the cache read price. Both are in addition to uncached input tokens.
- Batch API applies the model's published multiplier (50% on every current model) to every token line.
- Month = 30 days, year = 365 days. Taxes, volume discounts and provider commitments are excluded.
- Long-context tiers (Gemini above 200k input tokens) switch automatically when the input token count crosses the threshold.