LLM API Prices exists to make model pricing easier to verify and compare. It keeps official provider rates separate from OpenRouter rates, links published prices to their sources, and gives users a browser-based workload calculator without asking for prompts, API keys, or account access.
What problem does the site solve?
LLM price pages are difficult to compare directly. Providers publish different input, output, cache, batch, region, modality, and long-context rules. Routing platforms can expose another rate for the same model name. A simple “cheapest model” list can therefore hide which channel owns the price or whether two rows describe the same product.
LLM API Prices normalizes the standard text-token rates used in its catalog to USD per one million tokens. Each model keeps its provider, context limit, pricing status, source URL, verification date, lifecycle status, and channel-specific rates. The calculator then applies one workload consistently so users can see the effect of input volume, generated output, caching, repeated runs, and optional OpenRouter credit fees.
What makes this reference independent?
Every current price needs evidence
Verified prices link to the provider documentation or the OpenRouter model record used for that channel. A missing source does not become a guessed price.
Official and routed rates stay separate
An OpenRouter price is useful, but it is not relabeled as an official OpenAI, Anthropic, Google, or other provider price.
Suspicious changes wait for review
Source failures preserve the last verified value. Large or structurally unusual changes do not overwrite the catalog automatically.
What does LLM API Prices not claim?
The site does not claim that models with similar prices have equivalent quality, speed, safety, context behavior, tool support, or data policies. It does not provide an SLA, resell API access, process user prompts, or act as a model provider. Price rankings are not recommendations to use a model without testing it on the intended workload.
Published figures are reference estimates, not invoices or binding quotes. Taxes, negotiated contracts, free quotas, batch discounts, regional uplifts, tools, storage, images, audio, priority processing, and special long-context rules may require a separate calculation. Users should confirm important decisions on the linked official source.
Who is the site for?
The primary audience is developers, technical founders, engineering managers, finance partners, and researchers comparing API operating costs. The catalog is designed for pre-purchase research and budgeting: estimate a workload, understand the pricing components, shortlist viable channels, and then test the real API under production conditions.
Corrections are welcome. If a provider changes a standard rate, launches a model, retires a model, or changes how caching is billed, send the exact page, model, channel, current value, expected value, and official source through the contact page.