Estimate how much an LLM or AI API costs per 1,000 tokens based on input and output pricing. Enter your token counts and rates to quickly calculate total request cost and normalized cost per 1,000 tokens.
LLM pricing is usually split into input tokens and output tokens, with each billed at a different rate. Because providers often publish prices per million tokens, converting that pricing into a cost per 1,000 tokens makes it easier to compare models and estimate the cost of smaller requests. This calculator works by multiplying your input token count by the input rate and your output token count by the output rate, then adding both amounts together. It also normalizes the total request cost into an effective cost per 1,000 combined tokens, which is useful when prompts and responses vary in size. Using a cost-per-1,000-token view helps with budgeting, model comparison, and forecasting application expenses at scale. It is especially helpful for teams building chatbots, agents, summarization tools, or content workflows where token usage changes from one request to the next.
It is the normalized price of using 1,000 tokens, based on the provider's published input and output rates. This makes smaller usage amounts easier to understand than pricing shown per million tokens.
Many AI providers charge one rate for prompt tokens and another for generated response tokens. Output tokens are often more expensive because generation is typically more computationally intensive.
Yes. As long as you know the provider's input and output pricing and your estimated token counts, you can use this calculator for OpenAI, Anthropic, Google, Mistral, or other API-based models.
In 2026, a common budgeting mistake is using only the posted input/output rates and ignoring cache pricing. Many APIs now bill cached prompt tokens at a much lower rate than fresh input tokens, so long system prompts, tool schemas, and repeated context can cost far less after the first call. When comparing models, calculate two scenarios: cold requests and warm requests with cache hits. If your app reuses prompts across sessions, the cheaper model on paper may actually be more expensive in production.