Admin
Sign in with [email protected] to view Conviction’s data pipelines.
Sign in
Data catalog
Trace each brand from raw pulls to reporting.
Loading catalog…
Experiment costs
What each run spent on model calls, from the call files saved under runs/.
| Run | Last call | Calls | Failed | API cost | Estimated cost | Subscription calls | Unpriced calls | Model time |
|---|
API cost is the dollar cost the API reported for each call (OpenRouter). Estimated cost is tokens times list price, for APIs that report only tokens (Anthropic, Gemini).
Subscription calls ran on the Claude Code or Codex plans and add no cost per call. Unpriced calls used an API with no price in our table, so their cost is not counted.
A call copied into a later run counts only in the run that made it. OpenRouter's 5.5% credit fee is not included. Update: uv run python -m conviction.costs --push. Last update: .