Estimating Gemini API Costs

Two valid costing patterns are common: static token-price calculators and pipeline-aware operation costing.

Approach A: Static Token-price Estimator (Standalone)

Best for scripts, experiments, and prototypes.

interface ModelPricing {
  inputPrice: number;  // per 1M
  outputPrice: number; // per 1M
}

function estimateCost(model: string, inputTokens: number, outputTokens: number, pricing: Record<string, ModelPricing>): number {
  const p = pricing[model];
  if (!p) throw new Error(`Unknown model: ${model}`);
  return (inputTokens / 1_000_000) * p.inputPrice + (outputTokens / 1_000_000) * p.outputPrice;
}

You can pair this with a per-model and per-operation tracker.

Approach B: Pipeline-aware Operation Costing

Best for production pipelines with mixed cost semantics.

• Selection/script/policy: token-based.
• Image generation: often logged as fixed per-grid/per-image operation cost.
• Audio generation: token-based, but pricing may come from DB/config (models table).

await db.prepare(`
  INSERT INTO cost_logs (video_id, log_type, model_id, input_tokens, output_tokens, cost)
  VALUES (?, ?, ?, ?, ?, ?)
`).bind(videoId, logType, modelId, inputTokens, outputTokens, cost).run();

Practical Baseline Table (Pipeline Example)

Budgeting Guidance

1. Use static estimation before execution.
2. Reconcile with actual usageMetadata and operation logs after execution.
3. Recompute aggregate total_cost from cost_logs to avoid drift.

Note on Pricing Drift

Model prices and billing semantics can change. Keep pricing in config/data, not hardcoded in many files.