Guides
AI costs, explained in plain English.
Short, practical guides to the questions that decide your AI bill — what a token really costs, when the API beats a subscription, and where the hidden spend hides. Each one is backed by a free calculator.
AI Cost Per User for SaaS: The Unit Economics
Compute the AI cost of one user, check the margin against your price, and stop power users from eating your profit.
Read guideAI Voice Agent Cost Per Minute: The Three Bills
A voice agent bills for three services at once. See how to estimate cost per minute and per 1,000 calls, and where the money actually goes.
Read guideAI vs Human Writer Cost: The Honest Comparison
AI generation is cents; the real cost is the human editing behind it. Work out your true cost per word and when a human writer still wins.
Read guideCursor vs Copilot vs Claude Code: Real Cost
Usage-based billing changed the math in 2026. Here's how to pick the cheapest coding plan for your actual volume.
Read guideGGUF Quantization Sizes Explained
Q4_K_M, Q5_K_M and Q8_0 each pack a different number of bits per weight, which is what actually sets your GGUF file size and VRAM need.
Read guideHow Many Tokens Are in a PDF, Doc or Slide Deck?
Estimate tokens for a document before you feed it to an API using the ~1.33 tokens/word rule, then turn the count into a dollar cost.
Read guideHow Much Am I Spending on AI Subscriptions?
Stack up every AI plan you pay for, spot the overlap, and work out what's safe to cancel.
Read guideHow Much Energy Does ChatGPT Use Per Prompt?
What one AI prompt really costs in energy, carbon and water, and why the published numbers vary so much.
Read guideHow Much You Save Switching AI Models
Moving from a flagship to a cheaper model can cut token costs by 80 to 95 percent, if the smaller model holds quality on your task.
Read guideHow Prompt Caching and Batching Cut API Costs
Prompt caching can cut repeated-token cost by up to ~90% and batch APIs knock ~50% off, and the two discounts stack on the right workloads.
Read guideHow to Spot AI-Written Text (and Why You Can't Be Sure)
The countable tells of AI writing, and the honest reasons no checker can give you a reliable verdict.
Read guideSelf-Hosting vs API — When Is It Cheaper?
Renting or owning GPUs only beats paying per token above a break-even volume most teams never hit, once idle time and ops are counted.
Read guideTokens Per Second Explained (With a Wait-Time Example)
What tokens per second means, how it converts to words per minute, and how to turn tok/s into a real wait time.
Read guideWhy AI Agents Cost So Much
Agents resend the whole conversation every turn, so token cost grows quadratically and a 20-turn loop costs far more than 20 single calls.
Read guideWhy Reasoning Models Cost More
Reasoning models bill you for hidden "thinking" tokens at the output rate, which is why they often cost several times your first estimate.
Read guideSkip the reading — just run the numbers
Every guide is backed by a free, in-browser calculator. Browse them all.
See all calculators