pricing

**Pricing** AI pricing models must balance value delivery with sustainable unit economics, considering compute costs, API pricing structures, and the challenges of scaling AI products profitably. Common pricing models: per-token (OpenAI-style—pay for input/output tokens), per-request/API call (simpler for customers), subscription tiers (predictable revenue, usage limits), and value-based (price based on outcome delivered). Unit economics: cost to serve each request (GPU compute, inference time, model size); must have positive margin at scale. Track cost-per-query and compare to revenue-per-query. Pass-through costs: underlying model API costs (if using external models) often passed through with markup; customers understand this model. Usage-based challenges: unpredictable customer bills, need for cost controls, and difficulty forecasting revenue. Hybrid models: base subscription plus usage overage; provides predictability with scalability. Freemium considerations: free tiers can drive adoption but must convert to paid; AI costs make generous free tiers expensive. Enterprise pricing: often annual contracts with committed usage; volume discounts for large customers. Monitor margins: AI costs can change (model improvements, infrastructure efficiency); regularly review pricing against costs. Pricing strategy significantly impacts both customer adoption and business sustainability.