AI Model API Pricing Table (Live): Input/Output Rates and Context Windows Compared

How to read this table

The live table below lists official API prices for the models we continuously benchmark. LLM APIs bill per token, with input (what you send) and output (what the model generates) priced separately — output rates are usually much higher. The context window caps how much material fits in one request; long-context models are not necessarily pricier per token, but filling the window multiplies input cost.

The data shares the same model registry that powers our benchmark: whatever models we test are the models whose prices are tracked here, updated when vendors reprice.

Common cost-control tactics

Vendor-neutral strategies: put long, stable system prompts behind the provider's prompt-caching mechanism (cached input is typically discounted heavily); route by difficulty — send classification and extraction to small models and reserve flagships for hard reasoning; use batch APIs for non-realtime workloads; and cap output length — output tokens cost more than input tokens.

Beyond price: read it against capability scores

Price alone misjudges models that are cheap but wrong. Read this table against the YZ Index leaderboard (real sandbox execution, rule-based scoring): pick the cheapest model within a capability tier, not the cheapest overall. Per-model pages provide score history.