Claude Code pricing: Pro vs Max vs API

Claude Code is not free. It ships as part of the paid Pro and Max consumer plans, and it is also billed on its own through the Anthropic API if you would rather pay per token than hold a subscription. Every price below is what claude.com/pricing, the Claude Platform docs and the Claude Help Center showed in visible text on 28 September 2026. I pay for Max 5x myself, $100 a month, since 10 July 2026.

Is Claude Code free?

No. The features table on claude.com/pricing lists Claude Code as a row across all four plans: it reads No under Free, and Yes under Pro, Max 5x and Max 20x. The Free plan gets you Claude in a browser or app, not Claude Code in a terminal. Pro is the cheapest plan that includes it.

Plans at a glance

Free and Pro prices come from claude.com/pricing, checked on 28 September 2026. The page's own line under the Max card: "Usage limits apply. Prices shown don’t include applicable tax."

claude.com/pricing does not split Max into two prices on its main card. It shows "From $100" and "Per month" once for the whole Max plan, then "Choose 5x or 20x more usage than Pro" to describe the two tiers. The split by tier is on Anthropic's own Max plan help article, which names both: Max 5x at $100 per month, Max 20x at $200 per month. That article adds that these are web subscription prices and that mobile pricing may vary by app platform.

Team and Enterprise plans are sold per seat; their prices sit on the Team & Enterprise tab of the same pricing page.

How the usage limits work

Claude Code runs off the same subscription as Claude itself. Anthropic's Claude Code help article says usage limits are shared across both: "all activity in both tools counts against the same usage limits." That includes IDE usage in VS Code, Cursor or a JetBrains IDE, which "counts toward the same usage limits shared across Claude and Claude Code." You get warning messages as you approach the ceiling.

What happens when you hit the wall depends on your plan. Anthropic lists four ways out for Pro users: upgrade to Max 5x, turn on usage credits to keep going past the included limit, switch to a Claude Console account and buy API credits, or wait for the reset. Max 5x users get the same four, with Max 20x as the upgrade. One setting changes this outcome entirely: if you have an ANTHROPIC_API_KEY environment variable set, Claude Code uses that key instead of your subscription, and you get billed at standard API rates rather than drawing on your plan's included usage. Check that setting first if a plan feels like it is burning faster than expected.

Paying per token: the API

The other way to run Claude Code is to skip the subscription and pay Anthropic's API rates directly. Usage on this rail is billed separately, "at standard API rates (distinct from Pro/Max Plan pricing)," in Anthropic's own words. The table below is the full model price list from the Claude Platform docs, in dollars per million tokens:

Batch processing takes 50 percent off both input and output tokens for jobs that do not need to run right away. Fast mode, still in research preview, speeds up output on Claude Opus 5.5 at $8 per million input tokens and $40 per million output tokens; it only runs on the first-party Claude API, not on Claude Platform on AWS or partner clouds. If repeat calls are eating this budget before you even get to a decision, memoizing identical LLM calls and cache-first design are the two levers worth reading next.

What I pay, and why I moved from Max 20x to Max 5x

I pay for Claude Max 5x, $100 a month, since 10 July 2026. Before that I ran Max 20x. Moving down roughly halved my weekly hours on the top model, so today I push bulk, repetitive work off Claude by default and keep Claude for the work that needs judgment.

$ claude --version
2.1.283 (Claude Code)

That's the version on my box, checked today. The plan change also taught me something the pricing page does not show: the 5-hour window does not track only token spend. I went through a postmortem in May 2026 after a usage spike I could not explain from the dollar numbers alone. My peak day on 25 April was $2,172 in token-equivalent spend across 20 sessions, with 1 one-shot run that day, and the 5-hour meter stayed fine. My peak day on 4 May was $1,709, actually less money, across 114 sessions with 74 one-shot runs, and the meter burned fast. This is what I saw happen on my own account, not a metering formula Anthropic has published: a lot of short, disconnected claude -p runs seem to eat the window faster than a few long sessions carrying the same total spend.

My own rule of thumb since then, for one-shot claude -p runs in a day: under 10 is fine, 10 to 30 gets noticeable, past 30 turns problematic, and past 70 has been fatal to the window on the days I tracked it.

Which plan fits which workload

• Chat only, no terminal work: Free costs nothing, but it will not run Claude Code at all.
• Light, occasional Claude Code use: Pro, from $17 to $20 a month, is the cheapest plan that includes it.
• Coding across several projects most of the day: Max, sized 5x or 20x by how often you hit the shared limit. I run Max 5x myself.
• Scheduled jobs or pipelines that fire many short, separate runs: the API. It bills per token, separately from a subscription, and my rule of thumb above is exactly the pattern to watch for before scripting claude -p in a loop.

Prices in rupees

At ₹95.88 per US dollar (open.er-api.com, rate last updated Sun, 27 Sep 2026), $100 is about ₹9,588.

Related

• The Claude Code playbook
• The Claude playbook
• Claude Code skills guide
• Install Claude Code on Linux
• Claude Code in India: My Daily Setup

Sources

• Plans & Pricing | Claude by Anthropic (accessed 28 Sep 2026)
• Pricing - Claude Platform Docs (accessed 28 Sep 2026)
• Use Claude Code with your Pro or Max plan | Claude Help Center (accessed 28 Sep 2026)
• What is the Max plan? | Claude Help Center (accessed 28 Sep 2026)
• exchangerate-api.com via open.er-api.com (accessed 28 Sep 2026)