AI token budgets and provider costs
Learn what a link’s token budget measures, how model prices affect your bill, and how to size a budget for a real task.
By ShareMyTokens ·
What a token measures
Models receive and produce information in units called tokens. A prompt, prior messages included as context, and a model’s reply can all contribute to usage. The same number of words can lead to different token counts depending on language, formatting, and the provider’s tokenizer. That makes a fixed number of messages a poor substitute for measuring an actual task.
A ShareMyTokens token budget sets an allowance for the link. It helps a host allocate model usage across guests or experiments. If eligible API access is enabled, calls through the API and use of the guest chat draw from that same allowance.
From tokens to a cost estimate
Provider costs depend on the model and how the provider charges for the usage. Some providers publish separate rates for input, cached input, and output. As a simplified example, if a model charged $2 per million input tokens and $8 per million output tokens, 10,000 input and 2,000 output tokens would cost $0.036 before any other fees. These hypothetical rates illustrate the calculation.
At those same sample rates, 2,000 input and 10,000 output tokens would cost $0.084 even though both examples use 12,000 tokens. Check the provider’s rates and billing dashboard if you need a monetary estimate. The ShareMyTokens plan cost and the provider’s model usage charges are separate.
Measure before allocating
Start with one representative run. Use the model, prompts, document sizes, and approximate reply lengths that guests will encounter. Record the link’s usage before and after. Repeat for a longer or more complex task if the pilot needs both types of activity.
Suppose a rehearsal consumes 2,000 tokens. Three similar runs would suggest 6,000 tokens before allowing for variation. Choose an allowance within your plan that fits the number of runs and how much variability you expect. For a cohort, individual links give each person a defined allocation. Check the provider’s own usage controls when you also need to manage account-level spend.