AI API cost calculators help developers and businesses estimate token charges before a prototype becomes a production bill. Choose an OpenAI, Google Gemini or Anthropic Claude calculator below, enter realistic input and output volumes, and compare several models before committing to an architecture.
OpenAI, Gemini and Claude pricing was last checked on 19 August 2026. Estimates are in USD and should always be verified against the provider’s official pricing page.
OpenAI API cost calculator
Estimate standard text-token costs for GPT-5.6 Sol, Terra and Luna. Separate uncached input, cache writes, cache reads and output, with automatic long-context pricing above 272,000 input tokens.
Gemini API cost calculator
Estimate Standard or Batch text-token charges for current Gemini Flash and Flash-Lite models, including eligible cached input.
Claude API cost calculator
Estimate first-party Claude Fable, Opus, Sonnet and Haiku costs with Standard, Batch and supported Fast pricing, separate prompt-cache rates, US-only inference and successful web-search charges.
Which calculator should you use?
| Question | OpenAI tool | Gemini tool | Claude tool |
|---|---|---|---|
| Provider | OpenAI API | Google Gemini Developer API | Anthropic Claude API |
| Models included | GPT-5.6 Sol, Terra and Luna | Gemini 3.7 and 3.6 Flash, plus 3.5 Flash and Flash-Lite options | Active Fable, Opus, Sonnet, Haiku and limited-access Mythos models |
| Cached input | Included | Included | Separate 5-minute writes, 1-hour writes and cache reads |
| Batch mode | Not included in the current simple tool | Standard and Batch selectable | Standard and Batch selectable |
| Extra options | Core text-token estimate | Current promotional model rates | Supported Fast mode, US-only inference and web search |
| Output | Estimated USD token charge | Estimated USD token charge | Itemised USD estimate and per-request average |
Five inputs that make an API forecast useful
- A representative request. Measure a normal production prompt, including system instructions, conversation history, retrieved documents and structured data.
- A realistic output. Do not use the model’s maximum output length unless most requests actually reach it. Sample real or test responses.
- Monthly request volume. Multiply active users by requests per user, then include background jobs, evaluations and retries.
- Cache eligibility. Only put input in the cached field when your implementation and provider actually qualify it for the lower rate.
- A safety margin. Add headroom for growth, longer conversations, failed requests and seasonal peaks.
Why these tools use tokens instead of words
Modern API pricing is generally published per token or per million tokens, while a word-to-token ratio changes with language, punctuation, code, formatting and model tokenisation. A fixed “750 words equals 1,000 tokens” shortcut can be useful for rough thinking but is not accurate enough for production budgeting. Use actual token measurements from representative requests whenever possible.
Costs the calculators do not cover
- Provider tool charges not named by the selected calculator. The Claude tool includes successful web-search calls, but not every server-side tool.
- Images, audio, video, embeddings and fine-tuning.
- Cache storage and long-context surcharges not represented by the selected tool. The Claude calculator includes its documented US-only inference multiplier.
- Taxes, exchange rates and cloud infrastructure surrounding the API.
- Enterprise discounts, committed use or custom contracts.
A practical monthly budgeting method
Calculate a low, expected and high scenario. The low case can represent compact prompts and modest adoption. The expected case should use median production volumes. The high case should include peak traffic and longer conversations. Compare the same three scenarios across providers and models, then evaluate quality and reliability alongside price. The cheapest request is not economical if it fails the task and must be repeated.
Frequently asked questions
Are these official provider calculators?
No. They are independent planning tools using public provider pricing. Each calculator links directly to the relevant official source.
Do the estimates include free tiers?
No. The tools estimate listed paid token charges. Eligibility, limits and availability for free usage can change by account, region and model.
How often are prices checked?
The visible “last checked” date on each calculator records its latest review. Provider pricing can change between reviews, so confirm material budgets at the source.
For practical IT and endpoint-management guidance, return to the Lachie’s Lifestyle guide library.