Max effort in Claude’s iOS app now comes with a price label. According to a TestingCatalog sighting shared on 1 September 2026, the Claude app for iOS displays an explicit warning next to the Max effort option stating that it consumes 1.5 times the usage of a normal response. It is a small piece of interface text, but it is the first time Anthropic has put a number on what the deepest reasoning setting actually costs a subscriber — and it lands in the app where casual users are most likely to flip the switch without thinking.
This article follows our recent coverage of how AI vendors are exposing reasoning controls to users, including Perplexity’s granular effort selector and Anthropic’s unannounced Claude Hub orchestration surface. Here we explain what the leaked label says, how Claude’s five effort levels work, what a 1.5x multiplier does to your five-hour and weekly usage limits, and when Max effort is genuinely worth paying for. Everything is sourced from the published sighting and Anthropic’s own documentation, and we flag what remains unconfirmed.
Table of contents
- What the Claude iOS Max Effort Warning Actually Says
- How Claude’s Effort Levels Work, From Low to Max Effort
- Max Effort and Claude’s Usage Limits: The 1.5x Arithmetic
- Why Anthropic Added a Max Effort Warning Now
- Max Effort vs Rival Reasoning Controls
- Is Max Effort Worth 1.5x? A Decision Rule
- What the Max Effort Warning Means for Business Users
- What Remains Unconfirmed
- Max Effort Warning FAQ
- References
What the Claude iOS Max Effort Warning Actually Says
The sighting is brief and specific. TestingCatalog, the outlet that tracks unreleased and quietly shipped app features, posted two screenshots on Threads with the caption: “Claude app for iOS now has an explicit warning next to the Max effort option that it consumes 1.5 more usage.” The post closes with the shorthand “Max 1.5” and a warning emoji.
The leaked label, word for word
The screenshots show the effort picker inside the Claude iOS app, with the Max effort entry carrying a caution marker and a usage note. The only figure quoted is 1.5. There is no matching statement on Anthropic’s help pages at the time of writing, and the company’s own effort article says only that higher levels “take longer and use more tokens, so you’ll reach your usage limits faster” — without a multiplier.
What “1.5 more usage” most likely means
The natural reading is that a Max effort response is metered at 1.5 times the usage of the same response at the default level. That is consistent with how Anthropic describes usage: each message draws down a session allowance, and effort level is one of the listed factors that changes the draw. It is not consistent with a flat surcharge, and nothing in the sighting suggests Max effort costs extra money on a subscription.
Why iOS is the first place you would see it
Mobile is where impulsive settings changes happen. A desktop power user who selects Max effort has usually read the docs; a phone user scrolling a picker has not. Anthropic already notes that “mobile pricing may vary depending on your app platform”, so the iOS app has a separate commercial surface where a usage warning is easy to justify. Whether the same label has reached Android, web or desktop is not shown in the sighting.
How Claude's Effort Levels Work, From Low to Max Effort
Effort is Anthropic’s dial for how much work Claude puts into a response — thinking, text and tool calls together. Anthropic’s support article on effort settings lists five levels in the Claude apps, and the developer platform exposes the same five through an API parameter.
| Level | What Anthropic says it is for | Usage impact |
|---|---|---|
| Low | Routine tasks; “stretch your usage further” | Lowest |
| Medium | Routine tasks with moderate savings | Below default |
| High (Default) | “Best overall balance of quality and speed” | Baseline |
| Extra high (xhigh) | Long-running coding and agentic tasks | Above default |
| Max | “Deepest possible reasoning” | 1.5x on iOS, per the leaked label |
Effort is a signal, not a token budget
Anthropic’s platform documentation is careful on this point: “Effort is a behavioral signal, not a strict token budget.” At lower levels Claude will still think on hard problems, just less. At Max effort the model has “no constraints on token spending”, which is precisely why the usage cost is open-ended and why a fixed 1.5 label is notable — it suggests Anthropic meters the tier at a set multiplier rather than passing through the raw token count.
Which models offer Max effort
The support article lists effort controls as available for Opus 5, Sonnet 5, Fable 5, Opus 4.8, Opus 4.7, Opus 4.6 and Sonnet 4.6. The developer docs add that the xhigh step is newer and that “some models that support max don’t support xhigh”. If you want a refresher on how those models differ, our guide to the Claude 5 family covers the line-up.
Defaults differ by plan and product
In the consumer apps, High is marked as the default. In Claude Code, Anthropic’s power-user guide states the default is “high on Team, Enterprise, and direct API access, and medium on other plans”, adds an Auto level that picks per request, and warns plainly that “Max burns through usage limits faster, so activate it per session.” The iOS label is that same advice, rendered as a number and placed at the point of choice.
Max Effort and Claude's Usage Limits: The 1.5x Arithmetic
To understand why a 1.5x label matters, you need to know how Claude meters usage. Every paid plan runs two clocks. Anthropic’s Max plan article puts it directly: “Your session-based usage limit will reset every five hours”, and “Max plans also have a weekly usage limit that applies across all models.” The usage panel also shows a separate weekly reset for Opus alone.
What consumes usage
The usage limit best practices page lists the factors that increase draw: message length and attachments, the accumulated length of the conversation, tools such as Research and web search, the model selected, artifacts, and the effort level. Before this week, effort was the only item on that list with no published magnitude. Now, at least for Max effort on iOS, there is one.
What 1.5x does to a session
The arithmetic is simple and unforgiving. If a Max effort response draws 1.5 units against a default response’s 1.0, then the same session allowance funds two-thirds as many responses. A user who normally gets through a five-hour window comfortably will hit “Approaching 5-hour limit” — the warning Anthropic documents in its error-message guide — roughly a third sooner if every message runs at Max effort.
Turn the same figure around and it describes how far an allowance stretches. One hundred default-effort responses become sixty-seven at Max effort (100 divided by 1.5), so a third of the session simply disappears.
The plan tiers the multiplier lands on
The pain of a 1.5x draw depends entirely on the allowance behind it. Anthropic’s pricing page, checked on 1 September 2026, sets out the ladder below; the Max tiers are the only ones that advertise a usage multiple over Pro.
| Plan | Price | Usage position | Effect of running Max effort |
|---|---|---|---|
| Free | $0 | Baseline | Smallest allowance; limits arrive fastest |
| Pro | $20/month ($17 annual) | “More usage” than Free | Roughly a third fewer responses per session |
| Max 5x | $100/month | 5x Pro per session | Absorbs 1.5x easily; weekly cap still applies |
| Max 20x | $200/month | 20x Pro per session | Multiplier rarely binds; Opus weekly cap can |
| Team | $25/seat monthly ($20 annual); premium seat $125 ($100 annual) | “More usage than Pro” | Per-seat draw rises; admins see it in aggregate |
| Enterprise | $20/seat plus usage | Usage-priced | Cost scales with effort by design |
The irony is visible in the table: the warning is aimed at the subscribers who can least afford Max effort, yet the plan named Max is the one where the multiplier barely registers. A $20 Pro user on a phone is the person this label is really for.
Why Anthropic Added a Max Effort Warning Now
Interface warnings do not appear by accident. Three threads in Anthropic’s recent history explain the timing, and together they suggest the Max effort label is a considered piece of expectation management rather than a stray tooltip.
Effort control went mainstream in May
Until spring 2026 the effort dial was mostly a developer concern. That changed on 29 May 2026, when Anthropic launched Claude Opus 4.8 alongside a new effort selector available to all Claude users on claude.ai. Once every Free and Pro subscriber could reach Max effort from a menu, the question of what it costs them stopped being academic.
Usage complaints have been loud all year
Anthropic’s metering has been a sore point. A GitHub issue opened on 24 March 2026 collected reports from Max plan subscribers whose five-hour sessions were exhausted in one to two hours, with one Max 20x user describing a jump from 21% to 100% usage on a single prompt. The issue was closed as unrelated to Claude Code, but the underlying grievance — that heavy settings drain allowances invisibly — is exactly what a Max effort label addresses.
A wider push towards usage transparency
The label also fits a pattern. On 9 July 2026 Anthropic added a usage reflection dashboard for Free, Pro and Max users, with quiet hours and break nudges. The usage panel already shows session and weekly progress. Printing the Max effort multiplier next to the option is the logical next step: move the cost information from a settings page to the moment of decision.
Max Effort vs Rival Reasoning Controls
Every major assistant now offers some version of a thinking dial, and the interesting comparison is not the number of levels but whether the user is told the price. On that measure, the Claude iOS Max effort label appears to be a first among consumer apps — though rivals are converging on the same problem from different directions.
| Product | Control | Levels | Cost signal shown to the user |
|---|---|---|---|
| Claude apps (iOS) | Effort picker | Low, Medium, High, xhigh, Max | Explicit “1.5x usage” warning on Max effort (leaked, 1 Sep 2026) |
| ChatGPT | Thinking-time picker | Light, Standard, Extended, Heavy (Light and Heavy on Pro) | Speed described; no usage multiplier |
| Gemini app | Model choice + Deep Think (Ultra) | Compute-based quota, refreshes every 5 hours | Notifies when “close to your limit”; no per-mode figure |
| Perplexity Computer | Effort selector (in testing) | “Granular”, unannounced | Unknown; credits-based product |
| Claude API | effort parameter | low, medium, high, xhigh, max | Billed per token; effort shapes all tokens |
The industry converged on a dial
OpenAI’s consumer picker gives ChatGPT users Standard and Extended, with Light and Heavy reserved for Pro. Google’s Gemini apps moved to compute-based limits that weigh “the complexity of your prompt, the models and features you use, and the length of your chat”, with Deep Think restricted to AI Ultra. Each vendor has accepted that a large language model should not run at full depth on every question — a routine natural language processing request does not need frontier reasoning.
Only Anthropic prints the price on the dial
What none of the rivals do, on current evidence, is attach a number to the top setting. ChatGPT describes Heavy by speed; Gemini warns when you are near the limit but not what a given mode costs. A visible 1.5x on Max effort is a different design philosophy: tell the user before the draw, not after. If it proves popular, expect the others to copy it — the pattern in this market is that transparency features spread within a quarter.
Is Max Effort Worth 1.5x? A Decision Rule
A cost only matters against a benefit, so the practical question is when the top level earns its multiplier. Anthropic’s own guidance, and independent testing, point the same way: less often than most people assume.
What Anthropic itself recommends
The platform docs are unusually blunt for vendor documentation. For Opus 4.7 and 4.8 they advise: “Reserve max for genuinely frontier problems. On most workloads max adds significant cost for relatively small quality gains, and on some structured-output or less intelligence-sensitive tasks it can lead to overthinking.” For the newest Fable 5 model, they note that lower settings “still perform well and often exceed xhigh performance on prior models”. The recommended starting point for coding is xhigh, not Max effort.
What independent testing found
A MindStudio analysis published on 9 July 2026 reached a harsher verdict for API-billed Claude Code sessions: Max effort responses were “4–6x more expensive per query”, took 30 to 90 seconds against 5 to 15 at default, and produced marginal gains — sometimes worse results — on well-specified tasks. Those figures describe token billing rather than the subscription metering behind the 1.5x label, but the direction is identical.
Where the case for Max effort is real
Anthropic’s Opus 4.5 announcement in November 2025 showed why effort exists at all: at medium effort the model matched Sonnet 4.5’s best score using 76% fewer output tokens, and at highest effort it beat that score by 4.3 points using 48% fewer. The gains from the top setting are real but narrow — multi-step reasoning, ambiguous briefs, hard architecture decisions. That is the shortlist for Max effort.
| Task | Recommended level | Why |
|---|---|---|
| Drafting, summaries, quick lookups | Low or Medium | Anthropic: routine tasks “stretch your usage further” |
| Everyday analysis and writing | High (default) | “Best overall balance of quality and speed” |
| Long coding or agentic sessions | xhigh | Deeper reasoning “without the full token cost of max” |
| Hard debugging, architecture, frontier problems | Max effort, per session | Only place the 1.5x draw is likely to pay back |
| Structured output, extraction, classification | Medium, never Max | Anthropic warns Max “can lead to overthinking” |
A rule you can actually follow
Treat Max effort as a per-task decision, not a default. Switch it on for the one problem that justifies it, then switch back — the same “activate it per session” advice Anthropic gives Claude Code users. On a Pro plan, a day spent at Max effort is a day with a third fewer answers; on a Max 20x plan, the constraint is more likely to be the weekly Opus cap than the multiplier.
What the Max Effort Warning Means for Business Users
For an individual, the label is a budgeting aid. For an organisation deploying Claude across a team, it is a governance signal — and a reminder that reasoning depth is now a cost line that staff control from their phones.
Team and Enterprise administrators
On Team plans, every seat has its own allowance, and a handful of users who leave Max effort on will hit limits and escalate to IT. The fix is policy, not procurement: publish a short internal guide on when each effort level is appropriate, mirroring the decision table above. On Enterprise, where usage cost “scales with model and task”, Max effort translates directly into invoice lines, so finance will want the same guidance. A structured AI strategy review is the right place to set those defaults.
API users have a sharper version of the same problem
Developers do not see a 1.5x label; they see a token bill. The platform docs note that effort “affects all tokens in the response”, including tool calls, and that changing effort between requests “does not preserve cached prefixes” — so toggling Max effort mid-conversation also throws away prompt-cache savings. Teams running agents at scale should hold effort constant per workload and treat Max effort as an exception path. Our cost optimisation practice sees this pattern often: the expensive setting is rarely the default, but it is often left on by accident.
Train the habit, not just the setting
Anthropic’s usage guidance offers the cheapest wins: content in projects “is cached and doesn’t count against your limits when reused”, and bundling related questions into one message costs less than a string of follow-ups. Pairing those habits with a clear rule on Max effort is worth more than any plan upgrade. The label on iOS makes the conversation easier to have, because the number is now on the screen rather than buried in a help centre article.
What Remains Unconfirmed
The sighting is a single Threads post with two screenshots, and honesty demands listing what it does not establish. Anthropic has not published the 1.5 figure anywhere, has not described how it is calculated, and has not said whether it applies to every model that offers Max effort or only some. The label may be a test cohort rather than a general rollout; TestingCatalog’s post does not say how widely it appears.
Nor does the sighting show whether the multiplier reaches Android, the web app or the desktop clients, or whether xhigh carries a warning of its own. It is also possible that the figure is rounded — Anthropic’s documentation elsewhere treats effort as a behavioural signal with open-ended token spend, which fits awkwardly with a single fixed multiple. If the company publishes an official breakdown, the arithmetic above should be redone against it.
What the evidence does support is narrow but useful: Anthropic is now willing to state usage costs at the point of choice, the figure it has chosen for Max effort is 1.5x, and the audience is mobile users on plans where that multiplier bites. We will update this piece if the label is documented officially or reaches other platforms.
Max Effort Warning FAQ
Does Max effort cost extra money on a Claude subscription?
Not on the evidence available. The label describes usage consumption, not a fee. On Pro, Max 5x and Max 20x plans, running Max effort draws your session and weekly allowances down faster — 1.5 times faster per response, according to the leaked iOS label — but does not add to the monthly price. On Enterprise plans that are usage-priced, and on the API, higher effort does raise the bill.
Is Max effort the same as extended thinking?
No. Effort governs how much total work Claude does across thinking, text and tool calls; thinking is one component. Anthropic’s docs say effort is the recommended way to control thinking depth where adaptive thinking is available, and that at lower effort Claude “can skip thinking entirely for simpler problems”.
Should I ever use Max effort on my phone?
Rarely. The iOS app is where quick questions live, and quick questions are the worst use of the top setting. Anthropic’s own advice is to use Max effort for frontier-level problems and to enable it per session. If you find yourself needing it routinely on mobile, the better move is usually a higher plan or a different workflow, not a permanent Max effort setting.
Will the 1.5x Max effort warning come to Android and desktop?
Unknown. The sighting shows iOS only. Anthropic typically rolls interface changes across platforms within weeks, and there is no technical reason the label would stay iOS-exclusive, but nothing has been announced.
References
Claude Help Center: Change the model, effort, and thinking settings
Claude Help Center: Claude Code power user tips
Claude Help Center: What is the Max plan?
Claude Help Center: Usage limit best practices
Claude Help Center: Troubleshoot Claude error messages
TestingCatalog: Anthropic launches Claude Opus 4.8 and new effort selector
TestingCatalog: Anthropic adds usage reflection dashboard to Claude for all users
GitHub: Claude Max plan session limits exhausted abnormally fast (issue #38335)
MindStudio: Claude Code effort levels — why Max mode often hurts more than it helps
Anthropic: Introducing Claude Opus 4.5
TechRadar: You can now toggle GPT-5’s thinking time for faster or smarter answers
Google: Gemini Apps limits and upgrades for Google AI subscribers
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.