## Relevant issues Related to #30344 / #30880 — those fixed `get_model_info` dropping arbitrary `_above_*` keys from an existing cost-map entry. This issue is a remaining gap on the **Router custom-pricing registration** path. ## Describe the bug When configuring custom tiered pricing on a DB/UI deployment, e.g.: - `input_cost_per_token_above_32k_tokens` - `output_cost_per_token_above_32k_tokens` - `cache_read_input_token_cost_above_32k_tokens` these fields: 1. Appear correctly in Dashboard `litellm_params` (because `LiteLLM_Params` has `extra="allow"`) 2. Do **not** appear in spend-log `model_map_information.model_map_value` 3. Are **not** applied during cost calculation (requests in the 32k–128k range are billed at the base `input_cost_per_token`) Meanwhile `*_above_128k_tokens` works, because it is declared on `CustomPricingLiteLLMParams`. ## Root cause (code) Router only copies fields from `CustomPricingLiteLLMParams.model_fields` into the cost map: https://github.com/BerriAI/litellm/blob/v1.92.0/litellm/router.py#L7988-L8007 ```python for field in CustomPricingLiteLLMParams.model_fields.keys(): field_value = deployment.litellm_params.get(field) if field_value is not No…