Claude Code cost calculator: which plan do you actually need?
Pro, Max 5× or Max 20× — the honest answer depends on how much you really use Claude Code. This Claude plan calculator turns your weekly hours and model mix into a straight recommendation: which plan covers you, whether you'll hit the weekly limits, and what the same work would cost on the API.
Prices verified August 21, 2026 · runs entirely in your browser · nothing you enter is sent anywhere
Your usage
Time actually spent prompting — not hours the editor sits open.
Recommended
Claude Max 5×$100/mo
Comfortably inside this plan's weekly limits. 5× Pro usage, and Opus unlocks in Claude Code.
Same work on the API
~$456/mo
Subscription saves you
~$356/mo
- Claude Pro$20/moNot enoughNo Opus or Fable access in Claude Code on this plan.
- Claude Team (Standard seat)$25/moNot enoughNeeds at least 2 seats.
- Claude Max 5×Pick$100/moFitsComfortably inside this plan's weekly limits.
- Claude Team (Premium seat)$125/moNot enoughNeeds at least 2 seats.
- Claude Max 20×$200/moFitsComfortably inside this plan's weekly limits.
- Claude EnterpriseCustomNot enoughNeeds at least 5 seats.
| Plan | Cost | Weekly limits |
|---|---|---|
| Claude Pro | $20/mo | Not enoughNo Opus or Fable access in Claude Code on this plan. |
| Claude Team (Standard seat) | $25/mo | Not enoughNeeds at least 2 seats. |
| Claude Max 5×Pick | $100/mo | FitsComfortably inside this plan's weekly limits. |
| Claude Team (Premium seat) | $125/mo | Not enoughNeeds at least 2 seats. |
| Claude Max 20× | $200/mo | FitsComfortably inside this plan's weekly limits. |
| Claude Enterprise | Custom | Not enoughNeeds at least 5 seats. |
Estimates, not a bill. Weekly limits are measured in tokens, so the hour ranges Anthropic publishes are wide — where you land depends on your codebase size, context length and settings. API figures assume prompt caching is in use.
Weekly Claude Code hours by plan
| Plan | Price | Sonnet-class hours / week | Opus-class hours / week |
|---|---|---|---|
| Claude Pro | $20 / month | ~40–80 h | No Opus in Claude Code |
| Claude Max 5× | $100 / month | ~140–280 h | ~15–35 h |
| Claude Max 20× | $200 / month | ~240–480 h | ~24–40 h |
| Claude Team — standard seat | $25 / seat / month (2 seat minimum) | ~50–100 h (≈1.25× Pro) | No Opus in Claude Code |
| Claude Team — premium seat | $125 / seat / month (2 seat minimum) | ~250–500 h (≈6.25× Pro) | Included, Max-tier allowance |
How these numbers were worked out
Anthropic does not publish an hours-per-week figure for any Claude plan: the real limits are counted in tokens, across a rolling five-hour session window and a weekly allowance. The ranges above combine Anthropic's own published plan multipliers — Max is described as 5× and 20× Pro usage — with community telemetry from usage tools such as ccusage that log what real sessions actually burn.
Where you land inside a range depends on three things: repository size, how deep your agentic runs go, and your model mix. A Haiku-heavy hour costs about a third of a Sonnet hour and an Opus hour about 1.7×, so an identical number of hours can differ fivefold in what it consumes. Treat every figure here as a planning range, not a guarantee — a heavy week on a large monorepo can exhaust the bottom of a range, and a light week of question-and-answer work can comfortably beat the top of it.
Claude API prices per million tokens
| Model | Input / 1M tokens | Output / 1M tokens | Notes |
|---|---|---|---|
| Haiku 4.5 | $1 | $5 | Cheapest; ~0.33× a Sonnet hour |
| Sonnet 5 | $3 | $15 | $2 / $10 introductory pricing ends August 31, 2026 |
| Opus 5 | $5 | $25 | ~1.67× a Sonnet hour |
| Fable 5 | $10 | $50 | ~3.33× a Sonnet hour |
These are pay-as-you-go list prices, and they are what the calculator above converts your hours into. Two adjustments matter in practice: cached input is read back at roughly 10% of the input price, which is why long agentic sessions cost far less than their raw token counts suggest, and Sonnet 5 is currently discounted to $2 / $10 under introductory pricing that ends on August 31, 2026.
Prices verified August 21, 2026
Frequently asked questions
Is the Claude $200 plan (Max 20x) worth it?
Claude Max 20× at $200 a month is worth it mainly for Opus-heavy work — specifically if you need more than the roughly 15–35 Opus hours a week where Max 5× runs out. On Sonnet alone almost nobody needs it: Max 5× already covers 140–280 Sonnet-class hours a week, far more than a full-time developer gets through. The practical test is whether you actually hit a weekly cap on Max 5×, because if you have never seen a limit warning the extra $100 buys capacity you will not reach. Either tier still beats metering the same work through the API, which passes $200 a month at around eight typical hours a week — the Pro vs Max comparison has the full breakdown.
Is Claude Code's $20 Pro plan enough for daily coding?
Claude Pro at $20 a month is enough for daily coding if your sessions stay inside roughly 40–80 hours of Sonnet-class work a week and you do not need Opus. That comfortably covers solo and part-time use: code review, small features, debugging and short agentic runs on a normal-sized repository. It runs out quickly if you leave long autonomous agents working across a large monorepo, because the cap is measured in tokens rather than wall-clock time. Pro also has no Opus access inside Claude Code at all, so if you want the strongest model for architecture work the entry point is Max 5×.
How many hours of Claude Code do Pro, Max 5x, and Max 20x give per week?
Pro gives roughly 40–80 hours of Sonnet-class Claude Code a week, Max 5× roughly 140–280 hours plus 15–35 Opus hours, and Max 20× roughly 240–480 hours plus 24–40 Opus hours. These are estimates rather than published guarantees: Anthropic caps usage in tokens and does not state an hour figure for any plan. Where you land inside a range depends on repository size, how many files each run pulls into context, and your model mix. A heavy agentic week on a large codebase burns the allowance roughly twice as fast as steady conversational work.
When do Claude Code limits reset — and how do the 5-hour session and weekly limits interact?
Claude Code enforces two limits at once: a rolling five-hour session window that resets five hours after your first message in that window, and a weekly allowance that resets on a fixed seven-day cycle. Hitting the five-hour limit pauses you only until that window rolls over, while hitting the weekly cap pauses you until the weekly reset — that is the one that can end a working week. Because the session window is rolling rather than aligned to midnight, starting a burst at 9am means that window closes at 2pm. In practice most people on Max plans only reach the weekly cap if they run long autonomous agents most days.
Do claude.ai web chats and Claude Code share the same usage limit?
Yes — on Pro and Max plans, claude.ai web chats and Claude Code draw from one shared usage pool. A long afternoon of chatting in the browser directly reduces what is left for coding sessions that evening, because both are billed against the same subscription allowance. There is no separate Claude Code quota you can spend independently of the web app. The only real way to isolate the two is to run Claude Code against a metered API key instead of your subscription.
How much more usage does Claude Max give over Pro?
Claude Max gives five times or twenty times Pro's usage depending on the tier — Max 5× at $100 a month and Max 20× at $200 a month. Those multipliers are Anthropic's own framing and they apply to the token allowance, not to any published hour count. In hours that works out to roughly 140–280 Sonnet-class hours a week on Max 5× against 40–80 on Pro. The difference that decides most upgrades is not volume but access: Pro has no Opus in Claude Code, and both Max tiers do.
Does the Claude Team standard seat include Claude Code?
Yes — both Claude Team seat types include Claude Code, but with very different capacity. A standard seat at $25 per user per month carries roughly 1.25× Pro's usage, while a premium seat at $125 per user per month carries roughly 6.25× Pro's. Team needs a minimum of two seats either way, so the smallest real bill is $50 a month at standard. For one developer working alone, an individual Max plan is usually still cheaper, because the cheapest premium-led Team bill is $150 against $100 for Max 5×.
Is the Claude API cheaper than a Pro or Max subscription?
The Claude API is cheaper than a subscription only for very light use — roughly an hour a week or less of active Claude Code. At about $6 of tokens per typical active hour, even four hours a week costs around $100 a month, which is exactly what Max 5× charges flat, and twenty hours a week runs past $500 against a $200 plan. Subscriptions win as soon as usage is steady, because the flat price caps your downside on an unusually heavy week. The API is worth paying for on control rather than price — separate billing per project, no pool shared with claude.ai chats, no weekly cap — and the Claude API pricing breakdown has the per-token math.
How much does Claude Code cost per developer per month?
Claude Code costs $20 to $200 per developer per month on a subscription, or a reported $150–250 per developer per month when metered through the API. Anthropic has publicly reported both that API-metered figure and an average of about $13 per developer per day for its own internal engineering use; those describe teams running Claude Code heavily and continuously, not occasional use. On a flat plan the same developer pays $100 on Max 5× or $200 on Max 20× with no overage risk at all. Which side is cheaper comes down entirely to weekly hours, which is what the calculator above is for.
Why does ccusage report more usage than my plan costs?
Tools like ccusage report what your sessions would have cost at API list prices, which is almost always far higher than a subscription actually charges you. Most tokens in an agentic Claude Code session are cache reads — re-reading context that is already cached — and Anthropic bills those at roughly 10% of the normal input price, while usage reports commonly value them at or near full list rate. Adjusting for that mix, real API-equivalent spend for a typical session tends to land somewhere around 25–35% of the headline usage value a tool displays; that adjustment is a reasoned estimate from published cache pricing, not an official Anthropic figure. So a $3,000 monthly ccusage total does not mean your $200 plan is saving you $2,800.
