model_id often appears on multiple providers at different token rates. Sort search results to find the cheapest host today.
Search hosts for the model
Filter to exact matches
Keep rows wheremodel_id equals your target (for example gpt-4o). Ignore partial name matches.
Pick the lowest prompt rate
Comparepricing.prompt across the filtered rows. The lowest value is the cheapest input host for that model in the current sync.
You can repeat the check on pricing.completion when output tokens dominate your bill.
Price a request on that host
Once you pickprovider_id, call POST /v1/traces/cost with your token counts to confirm the dollar total.