What the site compares
AI Model Matrix normalizes public LLM API metadata into a small set of decision metrics: prompt price, completion price, context window, release date, and popularity signal. Prices are shown in USD per 1 million tokens so models can be compared on the same unit.
Price normalization
Prompt and completion prices are stored separately. Most comparison pages use 1 million input tokens and 500 thousand output tokens as the default workload example. Workload-specific guides may use a different scenario that is shown on the page, such as output-heavy chatbot traffic. These estimates are not bills or guarantees; they are standardized scenarios for comparing model economics.
Missing or unstable metadata
Some model entries may have incomplete pricing, release, or context metadata. Models with unavailable pricing are kept out of cost-ranked lists when a numeric price is required. Popularity-only models can appear when popularity data exists but pricing metadata is not available.
Popularity and freshness
Popular models come from current popularity signals. New models are marked from the 20 most recent release timestamps in the current catalog. The footer shows pricing data freshness so readers can tell when the underlying catalog was last refreshed.
Verification
AI Model Matrix is a comparison aid, not a contract or billing source. Provider prices, availability, limits, and model names can change, so production decisions should be verified against the provider dashboard or agreement before deployment.