--- title: 21 Days of LLMs, Day 2: Tokens, Context Windows, and Why... description: Tokens are the fundamental unit of LLM billing, not words — roughly 4 characters or 3/4 of a word each. Different providers use different tokenizers, meaning... canonical: https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: 21 Days of LLMs, Day 2: Tokens, Context Windows, and Why Your API Bill Just Exploded | daily.dev og:description: Tokens are the fundamental unit of LLM billing, not words — roughly 4 characters or 3/4 of a word each. Different providers use different tokenizers, meaning... og:url: https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v og:image: https://api.daily.dev/og/posts/C4oLbHc6v.png og:image:alt: 21 Days of LLMs, Day 2: Tokens, Context Windows, and Why Your API Bill Just Exploded og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # 21 Days of LLMs, Day 2: Tokens, Context Windows, and Why Your API Bill Just Exploded **[Medium](https://daily.dev/sources/medium_js)** · 5 min read · 2 upvotes · 0 comments ## Summary Tokens are the fundamental unit of LLM billing, not words — roughly 4 characters or 3/4 of a word each. Different providers use different tokenizers, meaning the same text can cost significantly more on one platform than another. Context windows in 2026 frontier models reach 1M+ tokens, but stuffing them creates runaway costs. A real-world example shows how sending 30 documents as a single prompt cost $47 in one afternoon, while switching to a vector database retrieval approach cut costs by 96%. Key takeaways: count tokens before every API call, add cost-logging wrappers from day one, and remember that per-request costs multiply brutally at scale. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why does the same text produce more tokens on Claude Opus than on GPT models? Different LLM providers use different tokenizers, so token counts for identical text vary between them. Anthropic has stated that Claude Opus 4.7's tokenizer can produce up to 35% more tokens for the same text compared to older models, which directly increases API costs for equivalent input since providers charge per token, not per word. _Comparing tokenizer overhead across models matters when budgeting LLM API spend, a topic developers track on daily.dev._ ### How can I reduce LLM API costs when querying a large set of documents? Instead of concatenating entire documents into every prompt, chunk the documents, embed them, store them in a vector database, and retrieve only the relevant chunks per question. One practitioner reduced average input tokens per request from 85,000 to 3,000 this way, a 96% cost reduction, while maintaining the same answer quality. _Developers cutting bloated LLM API bills with retrieval-based approaches follow this kind of guidance on daily.dev._ ### Why do code snippets consume more LLM tokens than plain English text of similar length? Python code typically produces 1.5x to 2x more tokens per line than plain English text because of syntax characters, indentation, and variable names. This means token estimates based on word count alone underestimate the true cost for code-assistant applications, so cost calculations for coding tools need to account for this heavier tokenization. _Anyone estimating LLM costs for a code assistant tracks tokenization quirks like this on daily.dev._ ## Similar posts on daily.dev - [Why Your LLM Bill Is 3× What You Expected](https://daily.dev/posts/why-your-llm-bill-is-3-what-you-expected-5dsz5eo2p) · DigitalOcean Community · 0 upvotes · 0 comments - [How To Reduce LLM Token Costs by 70–90%](https://daily.dev/posts/how-to-reduce-llm-token-costs-by-70-90--lqaosovr2) · C\# Corner · 1 upvotes · 0 comments - [How to Reduce AI Costs: 20 LLM Token Optimization Tips \(With Examples\)](https://daily.dev/posts/how-to-reduce-ai-costs-20-llm-token-optimization-tips-with-examples--5pefr2g2r) · finout · 1 upvotes · 0 comments - [Context window in AI: why every token is a budget decision](https://daily.dev/posts/context-window-in-ai-why-every-token-is-a-budget-decision-reskx1u2b) · Redis · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#nlp](https://daily.dev/tags/nlp) [View this post on daily.dev](https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"21 Days of LLMs, Day 2: Tokens, Context Windows, and Why Your API Bill Just Exploded","url":"https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v"},"datePublished":"2026-07-13T01:28:00.747Z","dateModified":"2026-09-14T07:25:12.814Z","description":"Tokens are the fundamental unit of LLM billing, not words — roughly 4 characters or 3/4 of a word each. Different providers use different tokenizers, meaning...","image":"https://media.daily.dev/image/upload/s--P4t4XyoV--/f_auto/v1722860399/public/Placeholder%2001","thumbnailUrl":"https://media.daily.dev/image/upload/s--P4t4XyoV--/f_auto/v1722860399/public/Placeholder%2001","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,nlp","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"21 Days of LLMs, Day 2: Tokens, Context Windows, and Why Your API Bill Just Exploded"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/21-days-of-llms-day-2-tokens-context-windows-and-why-your-api-bill-just-exploded-c4olbhc6v#faq","mainEntity":[{"@type":"Question","name":"Why does the same text produce more tokens on Claude Opus than on GPT models?","acceptedAnswer":{"@type":"Answer","text":"Different LLM providers use different tokenizers, so token counts for identical text vary between them. Anthropic has stated that Claude Opus 4.7's tokenizer can produce up to 35% more tokens for the same text compared to older models, which directly increases API costs for equivalent input since providers charge per token, not per word. Comparing tokenizer overhead across models matters when budgeting LLM API spend, a topic developers track on daily.dev."}},{"@type":"Question","name":"How can I reduce LLM API costs when querying a large set of documents?","acceptedAnswer":{"@type":"Answer","text":"Instead of concatenating entire documents into every prompt, chunk the documents, embed them, store them in a vector database, and retrieve only the relevant chunks per question. One practitioner reduced average input tokens per request from 85,000 to 3,000 this way, a 96% cost reduction, while maintaining the same answer quality. Developers cutting bloated LLM API bills with retrieval-based approaches follow this kind of guidance on daily.dev."}},{"@type":"Question","name":"Why do code snippets consume more LLM tokens than plain English text of similar length?","acceptedAnswer":{"@type":"Answer","text":"Python code typically produces 1.5x to 2x more tokens per line than plain English text because of syntax characters, indentation, and variable names. This means token estimates based on word count alone underestimate the true cost for code-assistant applications, so cost calculations for coding tools need to account for this heavier tokenization. Anyone estimating LLM costs for a code assistant tracks tokenization quirks like this on daily.dev."}}]} ```