LLM Token Cost Calculator
Compare API pricing across 0+ language models from OpenAI, Anthropic, Google, Meta, Mistral, xAI, DeepSeek and Cohere. Includes cached-input pricing, batch-API discounts and reasoning-token math. Prices auto-refresh every 24 hours via OpenRouter.
Featured model pricing (per 1M tokens)
| Model | Provider | Input | Output | Cached input | Context |
|---|
Frequently asked questions
How are these LLM prices kept up to date?
Prices are pulled directly from OpenRouter’s public models API once every 24 hours via a Convex cron job, then normalized to per-1M-token figures. Each row displays the timestamp of the last successful refresh.
What are cached-input and batch-API discounts?
Cached input applies when you send the same system prompt or context repeatedly — providers like OpenAI and Anthropic charge 50–90% less for cached tokens. Batch mode processes requests asynchronously within 24 hours for a flat 50% discount on both input and output. Toggle either in the calculator above to model real-world cost.
How accurate are the token estimates?
The paste-text estimator uses a rough chars÷4 heuristic. For mission-critical billing forecasts, run your prompts through the official tokenizer for your provider (tiktoken for OpenAI, Anthropic’s count endpoint, Gemini’scount_tokens API).
Which model is cheapest for high-volume production use?
For raw price-per-token, the small-model tier (GPT-5 nano, Claude Haiku 4.5, Gemini 3 Flash, DeepSeek V3) is typically 10–50× cheaper than the flagship reasoning models. Use the calculator to compare your specific input/output ratio — output tokens are usually 3–5× more expensive than input tokens, so verbose models can lose to less-verbose ones at scale.