--- title: Claude Code's effort levels matter more than you think description: An Anthropic engineer's breakdown of Claude Code's four effort levels (Low, Medium, High, Max) shows dramatic time differences on the same task, from about one... canonical: https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Claude Code's effort levels matter more than you think | daily.dev og:description: An Anthropic engineer's breakdown of Claude Code's four effort levels (Low, Medium, High, Max) shows dramatic time differences on the same task, from about one... og:url: https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun og:image: https://api.daily.dev/og/posts/hGqjBbSun.png og:image:alt: Claude Code's effort levels matter more than you think 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. # Claude Code's effort levels matter more than you think **[Trends](https://daily.dev/sources/trends)** · 3 min read · 6 upvotes · 1 comments ## Summary An Anthropic engineer's breakdown of Claude Code's four effort levels (Low, Medium, High, Max) shows dramatic time differences on the same task, from about one minute on Low to 28 minutes on Max. Lydia Hallie, who works on Claude Code, notes Low is the intentional default in Projects since the main thread mostly coordinates sub-threads, making higher effort levels often wasted compute. She also mentions Sonnet 5.5 is now available in Project settings. Additional token-saving tips include limiting repo exploration, using CLAUDE.md for persistent rules instead of repeating prompts, and stripping images/noise from PDFs before feeding them to the model. The overall advice is to match effort level to task complexity rather than defaulting to Max or using the tool less overall. ## Content The token-thrift discourse around Claude Code is loud right now, and the sharpest data point is a cautionary one: someone copied Spotify's setup to save money and ended up paying more. ## The Spotify experiment backfired Spotify's open-source shunt plugin hands cheap, read-heavy work to a cheaper worker model. Spotify reportedly saw savings of up to 90% in its Java monorepo benchmarks. A tester ran it against Opus 5.5 with two workers: Qwen3-Coder-30B-A3B running locally via Lemonade Server, and Claude Haiku. Delegation used 23% more tokens with Qwen3-Coder and 34% more with Haiku. It also ran 7-10x slower than Opus alone. The reasons are mundane. Opus already reads files efficiently with grep instead of full reads. The plugin's hooks miss many of Opus's actual tool calls. And Opus still has to re-check or rewrite drafts that miss the spec. ## The effort dial is the new religion The loudest advice is about reasoning effort. @kloss_xyz is pushing an article from a Claude Code engineer on when to use Low, Medium, High, and Max. The headline claim: the same task takes 1 minute on Low and 28 on Max. His verdict: "Stop running Max on everything." @lydiahallie is asking Projects users why they bump the main chat's effort. Low is the default there because that chat mostly coordinates threads. You can override it in Project settings, and she notes Sonnet 5.5 is in there too. Another piece lands in the same place. Use Medium for routine work, keep persistent rules in CLAUDE.md and one-offs in prompts, and strip images and noise from PDFs. Its summary: use Claude more intentionally, not less. ## Your clever setup might be the problem A regular user audited their workflow against a beginner's and found their "advanced" habits were hurting: - Too many MCP servers caused tool-selection confusion and context bloat. - Memory layers piled up stale context that clashed with the actual repo. - A bloated CLAUDE.md hurt instruction adherence, against Anthropic's own advice to keep it short. - Plan mode on two-minute tasks just added a wasted round of thinking. The Spotify result fits the same pattern: sophistication isn't free. ## Tools to see where the tokens go Claude Code's new /cost command breaks usage down by model (input, output, cache reads, cache writes) with a dollar estimate, which isn't charged for Pro/Max subscribers. Subscribers also get attribution to skills, subagents, plugins, and MCP servers, useful for catching an unattended loop retrying the same fix. The advice: give clearer stopping points, use rewind to undo bad edits, and remember that compaction itself burns tokens. For iOS devs, there's a PreToolUse hook that pipes xcodebuild, swift build, and swift test output through xcsift. It compresses noisy Xcode logs while keeping errors, warnings, and test results. ## Questions this post answers ### What do Claude Code's Low, Medium, High, and Max effort levels actually mean for task time? They control how much compute Claude Code spends per task, with dramatic time differences: the same task that takes roughly a minute on Low can take about 28 minutes on Max. Low is the intentional default for the main chat thread in Projects because that thread mostly coordinates sub-threads, so raising effort there is usually wasted compute. _Developers tuning Claude Code costs can track workflow tips like this through daily.dev._ ### How can I reduce token usage when working with Claude Code on a large codebase? Limit how much of the repository Claude explores before editing, since it will otherwise read the entire codebase first. Store persistent instructions in a CLAUDE.md file instead of repeating them in every prompt, and strip images and decorative elements from PDFs before feeding them in, since the model doesn't need that extra content to understand the text. _Anyone managing Claude Code costs on large repos can follow practical tips like these on daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@akkitto** · 0 upvotes > What, if you over-engineer effort management and suddenly, it takes more time than it saves tokens? 🤔 ## Similar posts on daily.dev - [Claude Code Sends 4.7x More Tokens Than OpenCode Before Reading Your Prompt](https://daily.dev/posts/claude-code-sends-4-7x-more-tokens-than-opencode-before-reading-your-prompt-9m02iom1z) · Hacker News · 6 upvotes · 0 comments - [How We Cut Our Claude Code Token Usage 2.8x\!](https://daily.dev/posts/how-we-cut-our-claude-code-token-usage-2-8x--p2fzdwf2l) · Daily Dose of Data Science \| Avi Chawla \| Substack · 0 upvotes · 0 comments - [Claude Code wastes tokens on tasks that don't need them, but this one setting fixes it](https://daily.dev/posts/claude-code-wastes-tokens-on-tasks-that-don-t-need-them-but-this-one-setting-fixes-it-jlojr0p1q) · XDA Developers · 0 upvotes · 0 comments - [Spotify’s Backstage Portal cut my Claude Code token usage by 90%](https://daily.dev/posts/spotify-s-backstage-portal-cut-my-claude-code-token-usage-by-90--2beo3hwm6) · Spotify Labs · 4 upvotes · 1 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun) ```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":"Claude Code's effort levels matter more than you think","url":"https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun"},"datePublished":"2026-09-28T22:40:10.076Z","dateModified":"2026-10-05T19:37:57.367Z","description":"An Anthropic engineer's breakdown of Claude Code's four effort levels (Low, Medium, High, Max) shows dramatic time differences on the same task, from about one...","image":"https://pbs.twimg.com/media/HTVnJAoWYAAstdE.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HTVnJAoWYAAstdE.jpg","isAccessibleForFree":true,"articleSection":"Trends","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":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"llm,anthropic,claude-code,prompt-engineering","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Claude Code's effort levels matter more than you think"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun","comment":[{"@type":"Comment","text":"What, if you over-engineer effort management and suddenly, it takes more time than it saves tokens? 🤔","datePublished":"2026-09-29T07:31:09.001Z","url":"https://daily.dev/posts/hGqjBbSun#c-J3CMRIJtV","author":{"@type":"Person","name":"Daniel","url":"https://daily.dev/akkitto","image":"https://media.daily.dev/image/upload/s--FtwJqX4c--/f_auto/v1754900041/avatars/avatar_29TCpY2hJR72V3BlxPXzX?_a=BAMClqZW0"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/claude-code-s-effort-levels-matter-more-than-you-think-hgqjbbsun#faq","mainEntity":[{"@type":"Question","name":"What do Claude Code's Low, Medium, High, and Max effort levels actually mean for task time?","acceptedAnswer":{"@type":"Answer","text":"They control how much compute Claude Code spends per task, with dramatic time differences: the same task that takes roughly a minute on Low can take about 28 minutes on Max. Low is the intentional default for the main chat thread in Projects because that thread mostly coordinates sub-threads, so raising effort there is usually wasted compute. Developers tuning Claude Code costs can track workflow tips like this through daily.dev."}},{"@type":"Question","name":"How can I reduce token usage when working with Claude Code on a large codebase?","acceptedAnswer":{"@type":"Answer","text":"Limit how much of the repository Claude explores before editing, since it will otherwise read the entire codebase first. Store persistent instructions in a CLAUDE.md file instead of repeating them in every prompt, and strip images and decorative elements from PDFs before feeding them in, since the model doesn't need that extra content to understand the text. Anyone managing Claude Code costs on large repos can follow practical tips like these on daily.dev."}}]} ```