Validated session value
Basksweep-passing + public-safe rows onlyLoading.
Equivalent cost of our public usage across providers
Validated public usage · most expensive → least expensiveUses the same aggregate input/output token counts from deduplicated, validated, successful, public-safe session rows and applies them to each model-specific rate card. This is a counterfactual API list-price estimate, not observed spend or guaranteed savings.
Waiting for validated public usage evidence.
Comparable-workload calculator
Change workload and time assumptions · most expensive → least expensivePotential time value is not realized savings. Human time and API-equivalent price are separate dimensions.
Time invested value$0
Potential time value$0
Net time scenario$0
Scenario basisUser-defined
Charts
Loading feed…
Provider price comparison
USD / million tokens · source-linked snapshot · most expensive → least expensiveAll providers receive identical input/output counts. Model pricing is not generic provider pricing. Comparison excludes caching discounts, batch, tools, search, taxes, subscription fees and negotiated rates unless stated. Lower price does not imply equivalent quality.
| Provider / model | Input / 1M | Output / 1M | Workload estimate | Basis / snapshot | Source |
|---|
Repository session value ledger
Only public-safe bounded metadata| Session / run | Provider / model | Input / output tokens | Outcome / validation | Elapsed | Paid equivalent | Evidence |
|---|
Method rules
- Paid-equivalent = input tokens × input rate + output tokens × output rate, divided by one million.
- Call it avoided cost only when the session is confirmed no-cost; unknown billing status remains unknown.
- Headline requires validation_status=VALIDATED, basksweep_status=PASS, and publication_status=PUBLIC_SAFE.
- Missing data is unknown, not zero. Failed, duplicate, unvalidated or low-confidence sessions cannot inflate value.
- Never publish prompts, completions, tool payloads, credentials, private quota readings or personal identifiers.