LLM API prices per 1M tokens

Every model sits in one table: input and output price per 1M tokens, context window and capabilities. Models in the catalog: 293, providers: 15. Sort the list by price, switch the currency, and add models to the comparison to see their specs side by side.

That is how providers publish them: a separate rate for a million input tokens and a million output tokens. A single unit makes models directly comparable with no math in between. For most models output costs more than input, so the bill depends heavily on answer length.

Provider price lists are denominated in dollars, and the ruble figures are a conversion at the official Bank of Russia rate, currently 1 $ = 84.51 ₽. The rate and the date it was taken are shown above the table and refresh together with the catalog. The currency switch restores the dollar prices.

The sync runs every day at 06:00 Moscow time and re-reads provider price lists, context limits, the model list and the Bank of Russia rate. A changed rate reaches the catalog at the next run, so there is no manual editing in between. Current data is as of September 2026.

Providers: 15, models: 293 — OpenAI, Anthropic, Google, Mistral, xAI, DeepSeek, Alibaba and other first-party model developers. Aggregators and resellers are excluded so the same model never shows up several times. Each provider also has its own page with prices.

For some models providers charge separately for cache reads and writes and for reasoning tokens — those rates appear on the model page whenever they are published. The shared table compares the base numbers: input price and output price per 1M tokens.

No. Foreign model developers in the catalog: 15, models: 293, all priced from their public price lists. Russian services, including GigaChat and YandexGPT, are not part of it — check their rates in the vendors' own documentation.