PrismHubUsage & token budgets
Turn per-request token usage into a useful picture of your workload.
Separate the token counts
Input usage can include instructions, conversation history, retrieved documents, and tool results. Output usage reflects the generated response under the selected model’s billing rules. Cache reads and writes are preserved as separate metering fields. Missing final metering stays pending instead of being estimated.
{
"usage": {
"prompt_tokens": 1500,
"completion_tokens": 500,
"total_tokens": 2000,
"prompt_tokens_details": {
"cached_tokens": 1000,
"cache_write_tokens": 0
}
}
}Calculate a text-workload estimate
PrismHub applies OpenRouter reference rates plus 5% to each category. Separate fresh input, cache reads, applicable cache writes, and output before multiplying by their respective rates. Cached tokens are not charged again as fresh input.
Keep the rate context
Store the model version, applicable rates, and relevant provider billing rules alongside your cost calculation. A generic rate cannot account for cached inputs, images, audio, or other separately billed usage.
Measure more than averages
Monitor long requests and multi-step workflows alongside typical conversations. Use per-request output limits and application budgets to catch outliers before they dominate usage.
Open the workload calculator