Claude Code Pricing Explained: Pro vs Max vs API

Claude Code Pricing Explained: Pro vs Max vs API

Claude Code has become one of the most capable AI coding CLIs available, but its pricing structure trips people up. Between subscription tiers, rolling rate limits, weekly caps, and a completely separate API pricing model, it's easy to either overspend or underbuy. This guide breaks down exactly what you get at each level as of February 2026, following the release of Opus 4.6 and the Agent Teams feature on February 5.

If you're comparing Claude Code's pricing against Gemini CLI or Codex CLI, check out our full comparison guide or the complete pricing guide that covers all three tools.

The Subscription Tiers

Claude Code is available across four tiers. Each tier determines your rate limits -- how much you can use the tool before getting throttled -- rather than unlocking different features. Every tier gets the same Claude Code capabilities, including access to Opus 4.6 and Agent Teams.

Free Tier

Cost: $0/month

The free tier gives you basic, light-use access to Claude Code. Think of it as a way to try the tool and understand how it works before committing. You'll hit rate limits quickly during any sustained coding session, which makes it impractical for daily professional use. But for occasional questions, small one-off tasks, or simply evaluating whether Claude Code fits your workflow, it costs nothing to find out.

Pro ($20/month)

Cost: $20/month

Pro is where Claude Code becomes a genuine daily driver. You get 5x the rate limits of the Free tier, which translates to roughly 40-80 hours of Claude Code usage per week depending on the complexity of your tasks and how token-intensive your conversations are.

For most individual developers, Pro is the sweet spot. A typical day of coding involves bursts of AI interaction interspersed with thinking, reviewing, testing, and meetings. Pro's limits accommodate that pattern comfortably. You'll only feel the limits if you're running Claude Code continuously for hours without breaks -- which, realistically, describes heavy refactoring days or large feature implementations.

The token budget works out to approximately 44,000 tokens per 5-hour rolling window. We'll explain what that means in the rate limits section below.

Max 5x ($100/month)

Cost: $100/month

Max 5x gives you 5x the Pro limits (25x Free). The per-window budget jumps to roughly 88,000 tokens per 5-hour window.

This tier exists for developers who regularly run into Pro limits -- typically those using Claude Code for extended autonomous sessions, Agent Teams workflows, or large-scale refactors that burn through tokens quickly. If you're finding that you hit Pro's ceiling two or three times per week, Max 5x is the pragmatic upgrade.

Max 20x ($200/month)

Cost: $200/month

Max 20x is the top tier: 20x Pro limits (100x Free), with approximately 220,000 tokens per 5-hour window.

At this level, rate limits are effectively a non-issue for all but the most extreme usage patterns. This is the tier for developers who treat Claude Code as their primary coding partner throughout the entire workday, teams piloting Claude Code with heavy usage expectations, or anyone doing sustained Agent Teams work across multiple parallel sub-agents.

Tier Summary Table

Understanding the Rate Limit System

Claude Code's rate limits confuse more people than the pricing itself. Here's how the system actually works.

The Rolling 5-Hour Window

Instead of a simple daily or monthly cap, Claude Code uses a rolling 5-hour window. At any given moment, your available quota is determined by how many tokens you've consumed in the previous 5 hours.

Think of it like a rolling average rather than a hard reset. If you burn through a large chunk of your quota at 9 AM, those tokens start "falling off" the window at 2 PM, and you regain capacity gradually rather than all at once at midnight.

This is actually better than a daily cap for most workflows. It means a heavy morning session doesn't wipe out your entire day. By afternoon, you've recovered most of your budget. Conversely, it prevents you from blowing your entire daily budget in one massive session and being locked out for the remaining 19 hours.

Weekly Caps

On top of the rolling window, there are weekly caps that prevent sustained maximum-rate usage from exceeding what the tier is designed for. In practice, the weekly cap matters less than the rolling window for most users. You'd need to be running Claude Code at maximum intensity for many hours per day, every day, to hit the weekly ceiling before the rolling window becomes your natural throttle.

What Happens When You Hit Limits

When you hit a rate limit on Claude Code, the tool doesn't stop working -- it throttles. Your requests still go through, but you may experience:

• Longer response times
• Automatic model fallback (from Opus 4.6 to Sonnet 4.5 in some cases)
• Queuing during peak usage periods

This is a better experience than a hard cutoff. You can still get work done; it just gets slower. For more detail on how rate limits work across different AI coding tools, see our rate limits guide.

API Pricing: The Alternative Path

If subscriptions don't fit your usage pattern -- maybe you have extremely variable usage, or you're integrating Claude Code into CI/CD pipelines -- you can run Claude Code against Anthropic's API directly using your own API key.

Current API Pricing (February 2026)

When API Pricing Beats Subscriptions

The crossover math is straightforward. A Pro subscription at $20/month gives you roughly 40-80 hours of usage. If your API costs would exceed $20/month, the subscription is cheaper.

API pricing wins when:

• Your usage is highly variable (heavy some weeks, zero others)
• You're running Claude Code in automated pipelines where cost-per-task budgeting matters
• You want to mix models -- using Haiku 4.5 for simple tasks and Opus 4.6 only when you need maximum capability
• You need to track and allocate costs per project or client

Subscriptions win when:

• You use Claude Code consistently throughout the week
• You prefer predictable monthly costs
• You don't want to think about per-token economics while working
• Your usage fits within the tier's rate limits

For a practical example: a moderately busy day of Claude Code usage (say, 3-4 hours of active interaction with Opus 4.6) might consume 150K input tokens and 50K output tokens. That's roughly $2.00 in API costs. If you have 20 such days per month, you're at $40 -- meaning Pro ($20) is a better deal, but Max 5x ($100) only makes sense if you're regularly exceeding Pro's rate limits.

Opus 4.6 and Agent Teams: What Changed

On February 5, 2026, Anthropic released Opus 4.6 alongside the Agent Teams feature. Both are available across all Claude Code tiers (including Free, though rate limits make sustained use impractical at that level). Here's why this matters for pricing decisions.

Opus 4.6 Performance

Opus 4.6 is measurably better than its predecessor at multi-step coding tasks. In our testing, first-pass correctness improved from roughly 90% to approximately 95% on standard development tasks. That 5-point improvement translates directly to fewer revision cycles, which means fewer tokens consumed per completed task.

The practical impact on pricing: Opus 4.6's higher quality means you use fewer tokens per task on average, which effectively stretches your rate limit budget further. A Pro subscriber on Opus 4.6 gets more done per window than they would have on the previous model.

Agent Teams Token Usage

Agent Teams is the more significant factor for pricing. When Claude Code spawns sub-agents, each sub-agent consumes tokens from your rate limit budget. A three-agent team working on a refactoring task uses tokens roughly three times as fast as a single-agent session.

This is the primary reason Max 5x and Max 20x exist. If you're using Agent Teams regularly -- and for complex tasks, you should be -- Pro's token budget gets consumed significantly faster than single-agent work. The orchestrator agent adds overhead on top of the sub-agents' token usage for coordination and conflict resolution.

Rule of thumb: If you plan to use Agent Teams as your default working mode, start at Max 5x ($100/mo) minimum. Pro will feel restrictive within the first day.

Team and Enterprise Considerations

Claude Code's subscription model is individual -- each developer needs their own subscription. For teams, this creates a straightforward but potentially expensive math problem.

The Team Cost Calculation

For larger teams, API pricing with organizational billing often makes more sense than individual subscriptions. You get centralized cost tracking, per-project allocation, and the flexibility to set different spending limits for different team members or projects.

Mixed-Tier Strategy

Not every developer on your team needs the same tier. A practical approach we've seen work well:

• Senior engineers and tech leads on Max 5x or Max 20x (heavy usage, Agent Teams, complex tasks)
• Mid-level developers on Pro (regular daily use, occasional rate limit hits)
• Junior developers and part-time contributors on Free or Pro (learning the tool, lighter usage patterns)

This mixed approach can cut team costs by 40-60% compared to putting everyone on the same tier. For more on managing AI tool costs across a team, see our engineering manager workflow guide.

Which Tier Is Right for You?

Here's a decision framework based on what we've seen across engineering teams we've worked with:

Start with Pro ($20/mo) if:

• You're an individual developer adopting Claude Code for the first time
• You code 4-8 hours per day with AI assistance interspersed (not continuous)
• You primarily use Claude Code for task-at-a-time work rather than long autonomous sessions

Move to Max 5x ($100/mo) if:

• You regularly hit Pro rate limits (you'll know -- the throttling is noticeable)
• You use Agent Teams for multi-file tasks that spawn multiple sub-agents
• You've measured the time cost of throttling and it exceeds $80/month in lost productivity

Go Max 20x ($200/mo) if:

• You treat Claude Code as your primary development environment
• You run sustained Agent Teams sessions that span hours
• Rate limits at any lower tier are a consistent daily frustration
• You're an engineering manager evaluating Claude Code for team deployment (see our engineering manager workflow guide)

Use API pricing if:

• Your usage is unpredictable or project-based
• You need per-project cost tracking
• You want to use multiple models strategically
• You're running Claude Code in automation (CI/CD, batch processing)

Tips for Maximizing Your Quota

Regardless of which tier you're on, these practices help you get more out of your token budget:

1. Be Specific in Your Prompts

Vague prompts lead to longer conversations with more back-and-forth, which eats tokens. Instead of "fix the login flow," try "the /api/auth/login endpoint returns 500 when the email field is missing -- add input validation and return a 400 with a descriptive error message." The second prompt often resolves in one shot.

2. Use Context Wisely

Claude Code reads files you reference, and those count against your token budget. Instead of loading your entire codebase into context, point Claude at the specific files relevant to the task. The @file syntax lets you be surgical about what context gets loaded.

3. Batch Related Tasks

Rather than five separate conversations for five related changes, describe them all in one conversation. Claude Code maintains context within a session, so the second through fifth tasks benefit from context already loaded for the first. This is particularly effective with Agent Teams, which can parallelize the execution.

4. Drop Down to Sonnet for Simple Tasks

If you're on API pricing, not every task needs Opus 4.6. Code formatting, simple refactors, boilerplate generation, and straightforward bug fixes can be handled by Sonnet 4.5 at 60% of the cost, or Haiku 4.5 at 20%.

5. Monitor Your Usage

Claude Code shows your current rate limit status. Pay attention to it. If you're at 80% of your rolling window budget, consider whether the next task is urgent or can wait an hour for tokens to free up. This is especially relevant on Pro, where the window is tighter.

For more strategies on getting the most out of AI coding CLIs regardless of which tool you use, see our best practices guide.

Claude Code vs Alternatives: Is It Worth the Premium?

Claude Code isn't the only option, and it's not always the cheapest. How does the value compare?

Vs Gemini CLI's free tier: If budget is your primary constraint, Gemini CLI at $0/month delivers real capability. The model quality gap between Flash and Opus 4.6 is significant, but for many tasks Flash is sufficient. Start with Gemini CLI free, and upgrade to Claude Code Pro when you need the quality boost. See our Gemini CLI Free Tier Guide for details.

Vs Codex CLI (via ChatGPT Plus at $20/mo): At the same $20/month price point, Claude Code Pro and Codex CLI via ChatGPT Plus offer different value propositions. Claude Code has higher first-pass correctness and Agent Teams. Codex CLI has sandboxed execution and comes bundled with ChatGPT access. If you already pay for ChatGPT, Codex is effectively free. If you're choosing one or the other, Claude Code Pro delivers more autonomous coding capability.

Vs GitHub Copilot ($10-39/mo): Different tools entirely. Copilot is an IDE-integrated autocomplete and chat tool. Claude Code is an agentic terminal tool. They complement each other rather than compete directly. Many developers run both. For a detailed comparison, see our Claude Code vs Cursor vs Copilot guide.

The Bottom Line

Claude Code's pricing is competitive for what it delivers. At $20/month, Pro offers strong value for individual developers. At $200/month, Max 20x is expensive but defensible for developers whose time is worth significantly more than the subscription cost -- which, for most professional developers, it is.

The rate limit system (rolling 5-hour windows plus weekly caps) is more nuanced than simple daily limits but ultimately more developer-friendly. It prevents feast-or-famine usage patterns and ensures you have capacity available throughout the workday.

If you're still evaluating whether Claude Code is the right tool at all, our comparison of Claude Code vs Cursor vs Copilot covers the broader landscape, and the Gemini vs Claude vs Codex comparison focuses specifically on CLI tools.