Real-time token costs and subscription pricing across the AI ecosystem. Compare β models from β providers.
New Bring your own key β try the BYOK keyring demo β| Model | Provider | Released | Context | Input $/1M | Output $/1M | Modalities |
|---|---|---|---|---|---|---|
Loading models⦠| ||||||
GPT-4o costs $2.50 per 1M input tokens and $10.0 per 1M output tokens. These prices are sourced from OpenRouter and updated hourly. Always verify with OpenAI's official pricing page before billing decisions.
Claude 3.5 Sonnet from Anthropic costs varies per 1M input tokens and varies per 1M output tokens. It supports a 200K context window, making it one of the best-value frontier models for long-document tasks.
A token is a chunk of text processed by an AI language model. In English, roughly 1 token equals 4 characters or 0.75 words. AI APIs charge separately for input tokens (your prompt) and output tokens (the model's response). Prices are quoted per 1 million tokens ($/1M). A typical ChatGPT-length exchange uses 300β500 tokens total.
DeepSeek models are significantly cheaper than most OpenAI equivalents. DeepSeek V3 costs $0.210 per 1M input tokens compared to GPT-4o at $2.50 per 1M input tokens. For many tasks, DeepSeek delivers comparable quality at a fraction of the cost.
Several models offer completely free API access ($0 per token), including models hosted on OpenRouter from Meta (Llama), Mistral, and others. Use the "Free" filter tab in the table above to see all currently free models. Availability of free tiers can change β check token.app regularly for the latest pricing.
token.app refreshes pricing data every hour by automatically fetching from OpenRouter and official provider pricing pages. The "Updated" timestamp at the top of the table shows when data was last refreshed. Prices can change without notice β always confirm critical pricing with the official provider.
Input tokens are the text you send to the model (your prompt, context, examples). Output tokens are the text the model generates in response. Output tokens typically cost 3β5Γ more than input tokens because generating text requires more compute than reading it. When optimising costs, reducing your prompt length and caching repeated context can significantly lower your bill.
As of 2025, several models support extremely large context windows. Google Gemini 1.5 Pro and 1.5 Flash support up to 2M tokens. Anthropic Claude models support up to 200K tokens. OpenAI GPT-4o supports 128K tokens. Larger context windows allow processing longer documents, conversations, and codebases in a single request.
The Rankings tab shows real-time usage leaderboards for AI models and agents, sourced from OpenRouter. You can filter by time period β 24H (daily), 7D (weekly), or 30D (monthly) β to see which models are trending and which AI-powered apps and agents are consuming the most tokens. Rankings update hourly.
token.app tracks real-time token pricing, subscription costs, and usage rankings across the AI ecosystem. We aggregate pricing data and usage leaderboards from OpenRouter and official provider pricing pages, refreshing every hour so you always see current rates. Coverage spans 344+ models from 57+ providers β including frontier labs like OpenAI, Anthropic, Google DeepMind, Meta AI, Mistral, DeepSeek, xAI, Qwen, NVIDIA, and Cohere, as well as dozens of fine-tuned and open-weight variants.
Every row in the pricing table shows the model's input cost and output cost per 1 million tokens, its context window size, and the modality types it supports (text, vision, audio, reasoning). The Rankings tab shows model and agent usage leaderboards with daily, weekly, and monthly token volume, so you can see which AI models and applications are trending right now. Prices reflect the listed API rate; enterprise or volume discounts may differ. For the most accurate billing information always check the provider's official pricing page. Data is provided by Measurable AI and is intended for research and comparison purposes.