LLM cost calculator
Estimate the cost of an LLM workload from three numbers: input tokens per request, output tokens per request, and how many requests you expect. Cost is (input ÷ 1M × input rate) + (output ÷ 1M × output rate), multiplied by request count.
Estimate a workload
Enter the shape of one request and how many you expect. Cost is
(input ÷ 1M × input rate) + (output ÷ 1M × output rate),
multiplied by requests.
Token counts are estimates: roughly 1.3 tokens per English word, and less efficient for code or non-English text. Availability and pricing vary by model and by account.
What determines LLM API cost?
The model you choose and the number of tokens in and out. Rates are quoted per million tokens and differ per model; output is normally priced above input, often by a factor of three or more.
How are input and output tokens priced?
Separately. Input covers everything you send — the system prompt, the conversation history, any retrieved context — and is charged on every call, so repeated context is a recurring cost rather than a one-off. Output covers what the model generates.
Why can two models with the same token count cost different amounts?
Because the per-million rate is a property of the model. The same 1,000 input and 500 output tokens can differ by more than tenfold between the largest and smallest model in a single family. This is why model selection usually beats prompt micro-optimisation as a cost lever.
Frequently asked questions
How do I calculate LLM token costs?
(input tokens ÷ 1,000,000 × input rate) + (output tokens ÷ 1,000,000 × output rate), then multiply by the number of requests. Input and output have different rates.
Can I estimate my AI API spend before switching?
Yes — that is what this calculator is for. Enter your typical input and output token counts and your expected request volume. Where Astrodyne holds a verified price and a published direct rate for a model, both are shown side by side.
How many tokens is a word?
Roughly one word is 1.3 tokens in English, so 1,000 words is about 1,300 tokens. Code and non-English text tokenize less efficiently, so treat this as an estimate rather than a rule.
Why does the calculator show pricing as unavailable?
Because Astrodyne publishes a price only once it is approved, and a comparison only when there is also a current published direct rate to compare against. Rather than showing an invented figure, the calculator states which inputs are missing.