# CostOfToken

> Compare LLM API pricing across OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen, Kimi and more. Input, cached and output cost per 1M tokens, normalized to USD and updated daily. Free public JSON API, no signup.

All prices are USD per 1,000,000 tokens, standard tier, normalized across
providers so they are directly comparable. Updated daily from each vendor's
published pricing page where one is machine readable, and from the OpenRouter
catalogue otherwise (those rows are marked `source_kind: "api"` — OpenRouter
is a reseller, so its rate can differ from the vendor's own).

Batch, Flex and Priority tiers are deliberately excluded: comparing one
vendor's batch rate against another's standard rate would make the table wrong.

## Data

- [Complete pricing table, markdown](https://www.costoftoken.com/llms-full.txt): every model in one document
- [Current prices, JSON](https://www.costoftoken.com/api/v1/prices): the full table as JSON
- [Provider list, JSON](https://www.costoftoken.com/api/v1/providers): providers and model counts
- [Price history, JSON](https://www.costoftoken.com/api/v1/history/{model_id}): historical price points for one model

The JSON API is free and needs no signup. It is rate limited to 60 requests
per hour per IP.

## Pricing by provider

- [Qwen (Alibaba) pricing](https://www.costoftoken.com/providers/alibaba): 49 models, input/cached/output cost per 1M tokens
- [Claude (Anthropic) pricing](https://www.costoftoken.com/providers/anthropic): 15 models, input/cached/output cost per 1M tokens
- [ERNIE (Baidu) pricing](https://www.costoftoken.com/providers/baidu): 1 models, input/cached/output cost per 1M tokens
- [Doubao (ByteDance) pricing](https://www.costoftoken.com/providers/bytedance): 1 models, input/cached/output cost per 1M tokens
- [DeepSeek (DeepSeek) pricing](https://www.costoftoken.com/providers/deepseek): 13 models, input/cached/output cost per 1M tokens
- [Gemini (Google) pricing](https://www.costoftoken.com/providers/google): 29 models, input/cached/output cost per 1M tokens
- [Kimi (Moonshot AI) pricing](https://www.costoftoken.com/providers/moonshot): 9 models, input/cached/output cost per 1M tokens
- [OpenAI (OpenAI) pricing](https://www.costoftoken.com/providers/openai): 73 models, input/cached/output cost per 1M tokens
- [Grok (xAI) pricing](https://www.costoftoken.com/providers/xai): 6 models, input/cached/output cost per 1M tokens
- [GLM (Zhipu AI) pricing](https://www.costoftoken.com/providers/zhipu): 20 models, input/cached/output cost per 1M tokens

## Notes for citation

- Quote the `updated_at` field with any price; these change frequently.
- A price of 0 means genuinely free, not unknown. Unknown is null.
- Cached input applies to repeated prompt prefixes and is typically ~10% of
  the standard input rate.
- Some models bill a higher rate above a long-context threshold; that is
  reported separately as `long_input` / `long_output`.