Credit-Based Pricing

Customers buy or receive a pool of credits that get consumed as they use the product. Credits abstract away the underlying compute cost into a single, predictable currency. The key tension: vendors must decide how fast credits deplete (too fast = sticker shock, too slow = margin erosion).

Companies Using This Model

• clay -- dual-currency credits (data credits for enrichment, action credits for orchestration)
• elevenlabs -- character-based credits for speech synthesis and voice cloning
• runway -- credits for AI video generation, allocated per plan tier
• autify -- credit consumption varies by AI model (Opus/Sonnet/Haiku) via Aximo AI tester

Credit-Adjacent Companies

These companies use credit/token pools as their value metric but are formally classified under different pricing models:

• bolt -- per-seat pricing with token-based consumption meter underneath
• lovable -- flat-rate pricing with monthly credit pool shared across unlimited users
• replit -- hybrid (seat + credits), Core plan includes $20/mo credit allocation

Patterns & Trends

Dual-currency systems are emerging. Clay's split between data credits and action credits lets them price enrichment (high cost) separately from orchestration (low cost). This is more sophisticated than a single credit pool.

Model-aware credit burn. Autify's Aximo charges different credit rates depending on which AI model the user selects (Opus vs Sonnet vs Haiku). This passes LLM cost variability directly to the customer.

Credit squeeze in progress. Several vendors in Q1 2026 cut credit allocations -- Replit dropped Core credits from $25 to $20/mo (-20%), Relay cut included credits by 60%. Credits are the easiest lever to pull when AI costs don't drop as fast as hoped.

Credits vs usage-based. Credit-based pricing differs from pure usage-based in that credits are pre-purchased or bundled with plans. The customer commits upfront. This gives vendors revenue predictability but creates a "use it or lose it" dynamic.