OpenAI Token Cost Calculator
Calculate OpenAI API token costs for every current GPT model, from GPT-5.6 Luna to GPT-5.6 Sol.
Usage
$0.15/M in · $0.075/M cached in · $0.60/M out · 128K context
Estimated cost
$13.50/month
Per request
$0.00045
Per day
$0.45
Per year
$164.25
Compare across models
| Model | Provider | Per month |
|---|---|---|
| OpenAI | $7.50 | |
| OpenAI | $13.50 | |
| OpenAI | $24.00 | |
| OpenAI | $36.00 | |
| OpenAI | $37.50 | |
| OpenAI | $37.50 | |
| $46.50 | ||
| $78.75 | ||
| Anthropic | $105.00 | |
| OpenAI | $180.00 | |
| OpenAI | $180.00 | |
| $180.00 | ||
| OpenAI | $187.50 | |
| Anthropic | $210.00 | |
| OpenAI | $225.00 | |
| OpenAI | $240.00 | |
| $240.00 | ||
| Anthropic | $315.00 | |
| OpenAI | $420.00 | |
| Anthropic | $525.00 | |
| Anthropic | $525.00 | |
| Anthropic | $525.00 | |
| Anthropic | $1,050.00 |
How this is calculated
Providers bill input and output tokens separately, per million tokens (MTok). Cached input tokens — a repeated prompt prefix the provider recognizes — are billed at a discounted rate.
cost = (input_tokens − cached_tokens) × input_rate
+ cached_tokens × cached_rate
+ output_tokens × output_ratePricing is verified against provider pricing pages as of 2026-09-01. AI and cloud pricing changes frequently — confirm the current rate on the provider's own pricing page before budgeting.
Frequently asked questions
How many tokens is my prompt?
As a rule of thumb, 1 token ≈ 4 characters ≈ ¾ of a word in English. For exact counts, use OpenAI's tokenizer or the `tiktoken` library — this calculator takes token counts as input rather than raw text.
What's the price difference between GPT-5.6 tiers?
GPT-5.6 Sol, Terra, and Luna are priced roughly 20x apart between the cheapest and most capable tier — Luna costs $0.20/$1.20 per million input/output tokens, Sol costs $4.00/$20.00.
Does OpenAI charge for cached tokens?
Yes, at a discount — repeated prompt prefixes (like a fixed system prompt) are automatically detected and billed at the cached input rate, roughly 10% of the standard input rate, with no code changes required.
Are older models like GPT-3.5 still cheaper?
GPT-3.5 Turbo remains inexpensive but is now a legacy model; GPT-5-nano and GPT-5.6 Luna are comparably priced with significantly better capability for most tasks.