Klint

Token Formula

Klint bills using weighted tokens, not raw provider tokens.

Formula

weighted_tokens = raw_input_tokens × input_weight
                + raw_output_tokens × output_weight

Each model has a single weight multiplier that combines input/output costs into one number for simplicity.

Example

GPT-5.5 Mini has a 0.3× multiplier:

  • Request uses 1,000 input + 500 output raw tokens
  • Weighted: (1,000 + 500) × 0.3 = 450 weighted tokens

Claude Opus at 1.4×:

  • Same raw usage → (1,500 × 1.4) = 2,100 weighted tokens

Why weighted?

Direct provider pricing varies by model, modality, and batch discounts. Klint's flat weighted system lets you switch models without recalculating budgets. Your dashboard shows both raw and weighted counts.