Pricing guide
Tokens, credits, and flat LLM API plans.
The cheapest API is not always the lowest token price. Compare predictability, model quality, routing, limits, and engineering time.
Token metering
Metered APIs usually charge separately for input and output tokens. They are attractive when traffic is predictable, prompts are compact, and you want model-by-model cost control.
Credits and allowances
Credit plans convert usage into a vendor-specific unit or cap. Read expiration, overage, and daily-limit terms carefully because the headline monthly price may not describe the usable workload.
Flat monthly plans
Flat plans trade granular unit billing for predictable spend. Check concurrency, rate, context, acceptable-use, provider availability, and whether unlimited means no billing quota or literally unbounded infrastructure.
A practical evaluation
Run a representative prompt set, measure response quality and latency, test failure behavior, and estimate the engineering time required to maintain direct integrations. Choose on total product cost, not a single token price.