export interface GeminiModelPricing { /** USD per 1M input tokens. */ inputPerMTok: number; /** USD per 1M output tokens. */ outputPerMTok: number; /** * Context-cache rate per 1M tokens for the cached prefix. Gemini's * context cache bills separately from regular input. Undefined = * charge at `inputPerMTok` conservatively. */ cachedInputPerMTok?: number; /** If published, max context window in tokens — useful for router long-context slot selection. */ maxContextTokens?: number; } /** * Pricing for Gemini models used as the long-context slot per decision #10. * 2026-04 Google AI pricing; revisit on publish. * * Decision #10: "skip Deep Think (Ultra-tier overkill); use 2.5 Flash as triage". * Flash is a fallback for under-budget review on very small diffs; 2.5 Pro is * the canonical long-context slot (>50K input tokens per router default). */ export declare const PRICING: Record; export interface UsageBreakdown { inputTokens: number; outputTokens: number; cachedInputTokens?: number; } export declare function actualCost(model: string, usage: UsageBreakdown): number; export declare function estimateCallCost(model: string, estimatedInputTokens: number, maxOutputTokens: number): number; //# sourceMappingURL=pricing.d.ts.map