At DW Corp, Claude Code is the tool behind our Vibe Coding workshops. It is also the question we hear most often from people getting started: "Which plan should I subscribe to, and which model should I use?" The confusion is understandable because the word "tier" appears in two different places. This guide organizes both concepts and helps you choose the right combination for your pace of work.
What Is Claude Code
Claude Code is Anthropic's coding agent. Unlike an autocomplete tool, it works within a loop that you control: it reads your repository, edits files, runs commands and tests, and can even open a pull request. You follow each step and approve—or reject—what it proposes. It started as a command-line interface (CLI) tool, but it now also runs in VS Code, a desktop app, and the web.
The important point for anyone getting started is that Claude Code does not replace the developer. It speeds up manual work and proposes solutions, while you retain judgment, review, and responsibility for the code that enters the project.
The Two Meanings of "Three Tiers"
When someone talks about "Claude Code's three tiers," they could be referring to two different things. The first is the subscription plan, which determines how much you can use it. The second is the AI model you choose within the tool, which determines how deeply it reasons. Let's look at both.
Subscription Tiers: Pro, Max 5x, and Max 20x
All individual paid plans include Claude Code. The main difference between them is usage capacity—how many requests you can make before reaching the limit of the usage window.
- Pro (around US$20/month on current plans): ideal for people who use Claude Code a few hours a week for focused, contained tasks. It is the best starting point.
- Max 5x (around US$100/month): roughly five times the usage of Pro. It makes sense for people who code with the agent every day.
- Max 20x (around US$200/month): roughly twenty times the usage of Pro. It is intended for intensive, sustained use, with long sessions and multiple projects.
Prices change over time, and monthly and annual billing differ, so treat these figures as a reference and check the current plans on the official pricing page at https://claude.com/pricing before deciding.
Model Tiers: Haiku, Sonnet, and Opus
Within Claude Code, you choose which model drives the agent. Think of them as gears:
- Haiku: the fastest and most economical. Great for simple, repetitive, or high-volume tasks where speed matters more than deep reasoning.
- Sonnet: the balanced option. It handles most everyday workloads with good quality and reasonable cost. For most work, start here.
- Opus: the deepest-reasoning option. It is worthwhile for difficult problems, large refactorings, or complex debugging, where thinking more carefully can save hours.
How to Choose the Right Combination
The practical rule is simple. Start with Pro and Sonnet: it covers most use cases without breaking the budget. If you frequently hit the usage limit, upgrade your subscription to Max 5x. If the tasks become difficult, switch the model to Opus for that specific session and return to Sonnet afterward. Reserve Max 20x for when Claude Code becomes a central part of your day.
Using It with Control: Lessons from Our Vibe Coding Workshops
In our Vibe Coding workshops, we insist on one principle: do not accept generated code blindly. Read the diff, understand the proposal, and run the tests. AI makes mistakes, and the cost of accepting an error without review is yours, not the AI's.
The second principle is token efficiency. Ask for well-defined tasks, provide enough context, and avoid redoing work. Choosing the right model for each task—Haiku for simple work, Opus only when needed—is not just a matter of cost: it is also about efficiency and sustainability, something we take seriously in our projects.
Conclusion
Claude Code's "three tiers" are actually two complementary choices: how much capacity you subscribe to—Pro, Max 5x, or Max 20x—and how much reasoning you turn on—Haiku, Sonnet, or Opus. Start small with Pro and Sonnet, use the tool with review, and scale up according to actual demand. That is how the tool becomes a productive partner rather than a dangerous autopilot.