Blog

Plain-English guides to what AI really costs: tokens, API prices, subscriptions, video and image generation, and practical ways to spend less.

What Is a Token? A Plain-English Guide for AI UsersWhat AI tokens are, how many your text uses, why other languages need more, and why tokens decide cost and quality.
How to Estimate Your AI API Bill Before You BuildA simple formula and a worked example for estimating monthly AI API cost, plus the multipliers most estimates miss.
Why Output Tokens Cost More (and 6 Ways to Use Fewer)Output tokens cost 2-6x more than input on current AI APIs. Why, the hidden cost of reasoning, and how to cut it.
Context Windows Explained: Why Long Chats Get Worse and Cost MoreWhat a context window is, why AI forgets things in long chats, why long chats cost more, and what to do instead.
Prompting in Korean, Japanese or Hindi? How to Cut the Token CostNon-English prompts use up to 1.8x the tokens of English. Measured ratios and five ways to pay less, starting with one click.
How Much Does One Minute of AI Video Cost?One minute of AI video on Veo, Kling, Grok, FLUX, Seedance and more, from $3 to $24 at list price, and why retries triple it.
AI Image Generation Cost per Image, ComparedPrice per image on GPT Image, Nano Banana, FLUX.2, Grok Imagine, Seedream and Runway, from $0.01 to $0.20 at 1K.
ChatGPT Plus or the API: Which Is Cheaper for You?When a $20 AI subscription beats paying per token, and when the API is far cheaper, with monthly costs by usage.
What Does an AI Coding Agent Really Cost per Task?Why Claude Code, Codex and Gemini CLI use millions of tokens per task, what that costs on the API, and when a plan is cheaper.
JSON vs YAML vs CSV: Which Data Format Uses the Fewest Tokens?We measured one table in 7 formats. Pretty JSON used 2.9x the tokens of CSV, XML 3.6x. Results, costs and what to use.
How Many Tokens Does Code Use? Indentation, Comments and MinificationMeasured: indentation is almost free, docstrings are not, minified JS is half the tokens. Where coding tokens really go.
Why Numbers, IDs, Dates and Emoji Use More Tokens Than You ThinkMeasured on GPT's tokenizer: numbers split into 3-digit chunks, a UUID is 18 tokens, one emoji can be 11. How to save.
How GPT's New Tokenizer Cut Costs in 40 Languagescl100k vs o200k measured in 41 languages: Malayalam -82%, Hindi -67%, Korean -42%. How tokenizers work and what is left.
Prompt Caching Explained: Cut Repeated Input Costs by Around 90%How prompt caching works, why prompt order matters, a worked cost example, and a checklist to make sure your cache hits.
Write Shorter Prompts Without Losing Quality: Measured Before and AfterTwo real prompts rewritten: 98 to 20 tokens and 122 to 35. What to cut, what to keep, and what it saves at scale.
Reasoning Models: When Thinking Is Worth Paying ForHidden reasoning tokens are billed as output and can make a request 10x pricier. When reasoning pays off and when it does not.
Batch APIs: Half-Price AI for Work That Can WaitHow batch processing works, what it saves on a 100,000-item job, which tasks fit, and tips to avoid wasted runs.
RAG vs Long Context: What It Costs to Ask Questions About Big DocumentsSending a 500-page manual with every question vs retrieving 8 passages: $0.67 vs $0.009 per question. When to use each.
AI Cost per User: How to Price an App Built on an LLMEstimate AI cost per user per month, plan for heavy users and free tiers, and the levers that bring the number down.
AI Cost Glossary: 30 Terms Explained in Plain LanguageTokens, tokenizers, cached input, batch pricing, reasoning tokens, context windows, RAG and more, explained simply.
Same prompt, 41 languages: GPT token cost comparedOne prompt in 41 languages on GPT's o200k tokenizer: from 1.03× (Chinese) to 2.44× the tokens of English.
Claude Code Cost per Month: Pro vs Max vs APIWhat Claude Code really costs per month on Pro, Max 5x, Max 20x or the API, with cost per task and when each one is cheapest.
Claude Code Usage Limits Explained: 5-Hour and Weekly CapsHow Claude Code's 5-hour and weekly limits work on Pro and Max, what changed in 2026, how to check usage and what to do at the limit.
How to Save Tokens in Claude Code: 9 Habits That MatterWhy Claude Code uses so many tokens and 9 practical ways to cut them, from /clear and /compact to a shorter CLAUDE.md.
ChatGPT Pro vs Claude Max: Which $100 or $200 Plan Gives More?ChatGPT Plus/Pro vs Claude Pro/Max compared: usage per dollar after the September 2026 changes, models, coding agents and which to pick.
GPT-6 API Pricing: Astra vs Sol vs Luna, and What It CostsGPT-6 Astra, Sol and Luna API prices per million tokens, real cost per request, comparison with Claude and Gemini, and which to use.
LLM API Pricing Compared: 22 Models From GPT-6 to Claude, Gemini and DeepSeekInput and output API prices of 22 models side by side, the real cost of 10,000 requests, the cheapest LLM APIs, and how to choose.
Tokens per Word, Measured: English and 11 Other LanguagesHow many tokens per word and per character on current GPT models: about 1.1-1.3 per English word, and the real numbers for other languages.
How to Count Tokens for GPT, Claude and GeminiWhy GPT, Claude and Gemini count tokens differently, how to get an exact count for each, and a quick way to estimate without code.
ChatGPT Pro 100 vs 200 vs 500: Which Plan Is Worth It?ChatGPT Pro now comes in $100, $200 and $500 tiers. Usage per dollar, the Pro 200 cut, Ultrafast, and when the API is cheaper.

Blog

In other languages