API Cost Calculator
Compare LLM API pricing before a prototype becomes a surprise invoice. Estimate per-request cost, monthly spend, cached prompt savings, and batch pricing across OpenAI, Claude, Gemini, DeepSeek, Groq, xAI, Mistral, and more.
Prompt, context, retrieved docs
Visible answer plus reasoning
Expected production volume
Compare model pricing
Search the rate card, filter by provider, then click a row to load it into the calculator.
Best Models by Use Case
Hand-picked recommendations for common workloads. Click any model name to load it into the calculator.
Best for Coding
Models that excel at code generation, debugging, and refactoring across multiple languages.
Impressive coding ability at a fraction of the cost of frontier models. ($0.14/$0.28 per 1M tokens)
How to read the bill
API pricing is simple until tools, caching, and reasoning tokens join the party. Input tokens are everything you send: system prompt, chat history, files, retrieved chunks, and tool schemas. Output tokens are what the model generates, including invisible thinking tokens when a provider bills for them.
For production estimates, test with real prompts and real documents. A pricing calculator gives you the shape of the spend; provider usage logs give you the final truth.
Check official pricing for OpenAIPractical model picks
Use the lowest blended price that still passes evals. Start with small Flash, Nano, Haiku, or OSS models.
Budget for tool calls and retries. A stronger model can be cheaper if it avoids failed loops.
Context window matters more than sticker price once retrieval chunks get large.
Compare monthly spend, not only per-token rates. Output length quietly dominates support bots.
Built With Care
“The fastest way to waste AI budget is to benchmark with tiny prompts and then deploy agents that carry twenty turns of history, tool schemas, and retrieved documents. Run your real prompt shape here before you choose a default model.”
API Cost Calculator & LLM Pricing Comparison
Use this API cost calculator to estimate LLM spend from real token counts, daily request volume, cached prompt discounts, and batch pricing. It is built for developers comparing OpenAI API pricing, Claude pricing, Gemini API costs, DeepSeek pricing, Groq inference, xAI Grok rates, Mistral models, and other language model APIs before shipping to production.
- 1Pick the model you plan to use, or search the pricing table and click any model row.
- 2Enter input tokens for prompts, chat history, retrieved context, files, and tool schemas.
- 3Enter output tokens for the model response, including reasoning tokens when the provider bills them as output.
- 4Add daily request volume to estimate monthly spend.
- 5Use cached input and batch pricing toggles only when your provider and workflow actually support those discounts.
- Per-request and monthly LLM API cost estimates
- Searchable model pricing table with provider filters
- Comparison model selector with monthly savings or premium
- Cached input and batch pricing toggles
- Workload presets for chat, agents, RAG, batch jobs, and classification
- Official pricing links for provider verification
Real Ways People Use This
Estimate LLM API costs before launch
Enter realistic input tokens, output tokens, and requests per day to see per-request and monthly API spend before production traffic arrives.
Compare OpenAI, Claude, Gemini, DeepSeek, and Groq
Filter the model pricing table by provider and compare current input/output token rates across major commercial and open-weight inference APIs.
Plan cached prompt and batch savings
Toggle cached input or batch pricing when your workload qualifies, then compare the monthly delta against your standard real-time estimate.
Choose a model for agents and RAG
Use the workload presets to model chat, agent, RAG, batch, and high-volume classification patterns with less hand-waving.
- Provider prices change often. Use this as a planning tool, then verify the final number on the official provider pricing page.
- Reasoning, tools, web search, image/video, and audio features may add charges that simple text token math does not fully capture.
- Batch pricing usually means asynchronous processing. It is useful for offline jobs, not interactive user flows.
- 1Use real prompt and output samples, not guesses.
- 2Include chat history, retrieval chunks, and tool schemas in input tokens.
- 3Estimate monthly volume before picking a default model.
- 4Run a small production log sample through provider usage dashboards after launch.