Custom Model Pricing
Upload custom model prices for AI Observability and Agent Mode using CSV, LiteLLM, or Bifrost JSON.
Upload custom model prices when you have negotiated rates, use a model proxy, or need pricing for models outside the shared catalog. The same rates apply to AI Observability cost calculations and Agent Mode usage and budgets on the selected backend. Models missing from your file continue to use shared prices.
Open Custom Model Pricing
As a workspace admin, go to Settings > AI & Agents > Cost Control and find Custom model pricing. Select the backend whose prices you want to configure if your workspace has multiple backends.

Prepare a Price File
Choose one of these formats. Files can be up to 10 MiB.
LiteLLM JSON
Download the LiteLLM model pricing file and save it as a .json file. You can upload this file directly, or edit its supported price fields to match your rates.
The file is a JSON object keyed by model ID, with litellm_provider identifying the provider.
Bifrost JSON
Download the Bifrost pricing datasheet and save it as a .json file. You can upload this file directly, or edit its supported price fields to match your rates.
The file is a JSON object keyed by model ID, with provider identifying the provider. Upload the pricing model map rather than a proxy configuration file.
CSV Template
Select Download CSV template in the Custom model pricing card.
Open the downloaded
model-prices.csvin a spreadsheet or text editor.Keep the header row and replace the sample row with your own provider, model, and prices. Add one row per model.
Save the file as CSV.
provider
Required. Provider ID shown on your AI observability spans or in the Agent mode model list.
model
Required. Model ID, including its exact casing.
input_cost_per_token
Required. Base input price in USD per token.
output_cost_per_token
Required. Base output price in USD per token.
cache_read_input_token_cost
Optional. Cache read price in USD per token.
cache_creation_input_token_cost
Optional. Cache write price in USD per token.
All prices must be nonnegative numbers. Use 0 for a free rate. Leave an optional cache price blank to use the shared cache rate when available.
If your provider quotes prices per million tokens, divide by 1,000,000. For example, $2 per million input tokens becomes 0.000002 USD per token.
Replace openai and my-custom-model with the provider and model IDs you use, and enter your own rates. Match the IDs shown on your AI observability spans, including model name casing.
Copy Agent Mode Model IDs
For your convenience, provider and model IDs used by enabled Agent mode models can be copied from the list in the same card. Expand Provider and model IDs for enabled Agent mode models, use the copy buttons beside the provider and model, and paste the values into the corresponding CSV columns.
Where several IDs match a model, they are checked in the displayed order and the first priced match is used.
Upload and Activate Prices
Select Upload prices.
Choose your CSV or JSON file. Upload starts automatically.
Wait for the saved prices to become active. Allow up to two minutes for both AI Observability and Agent Mode to update.
After activation, new spans and model calls use the custom rates. Historical costs stay unchanged. These rates affect calculations in groundcover; they do not change what your model provider charges.
If the file is invalid, the upload fails and your existing prices remain unchanged.
Replace, Download, or Remove Prices
Once a file is uploaded, the card shows its filename, model count, who uploaded it, and upload time.
Replace prices: choose another file to replace the entire custom price list. Include every model you want to keep custom prices for; models omitted from the replacement return to shared pricing.
Download file: download the original uploaded file to inspect or edit it.
Remove custom prices: restore shared pricing for the backend. Allow up to two minutes for the change to take effect. Historical costs stay unchanged.
Related
AI Observability — how groundcover tracks model costs.
Agent Mode Cost Management — configure spending limits and view usage.
Last updated
