Why does LLM observability pricing charge so much for normal telemetry?
When comparing s built for LLM apps, the pricing is hard to justify. Charging more for (the record of a request and response sent to a model) makes sense, since counting tokens, calculating model costs, and storing huge prompts and all take extra work.
But also generate plenty of ordinary data, like REST requests and database queries. That ordinary data ends up billed at the exact same rate as an LLM trace.
This pricing pushes teams to stop collecting non-LLM data just to save money. Then when something breaks, the context that would explain the is already gone, leaving teams to either work blind or spend far longer debugging the incident.
Key points
- Charging more for is reasonable given the extra work of ing, cost calculation, and storing large prompts/
- The core complaint: ordinary agent data like REST requests and database queries gets billed at the same rate as
- This pricing pressures teams to drop non-LLM data from to cut costs
- When incidents happen, the missing context makes root-cause debugging slower or impossible