Claude Code pricing in 2026: plans, API costs, and how limits work
What Claude Code costs on Claude Pro, Max and Ultra, what it costs on an API key, how the 5-hour and weekly limits work, and how to run it on a fixed monthly budget without them.
Claude Code has two ways to pay: a Claude subscription, or an API key. They are priced differently, limited differently, and suit different people. Here is each, using Anthropic's own pages as of November 4, 2026, and a third option that sits between them.
On a Claude subscription
Claude Code is included in Anthropic's paid individual plans (claude.com/pricing):
| Plan | Price | Usage |
|---|---|---|
| Pro | $20 a month, or $17 a month billed annually | "More usage than Free, every 5 hours and every week" |
| Max | $100 a month | "5x Pro usage every 5 hours and every week" |
| Ultra | $200 a month | "10x Pro usage every week"; no 5-hour limit for verified accounts |
Anthropic's help article on Claude Code explains the limits: Pro and Max have a five-hour session limit and a weekly limit, the limits are shared between Claude and Claude Code, and when you reach one you can wait for the reset, enable paid usage credits, or switch to API credits.
What the plans don't publish is a number. "5× Pro" is relative to Pro, and Pro's allowance is not given in tokens or dollars — Anthropic says how fast you reach a limit depends on the model, how long your chats are, and how large your files are.
On an API key
With an API key, Claude Code bills per token at the model's price and has no session or weekly windows. These are the list prices for the Claude models:
| Model | Input / 1M | Cached / 1M | Output / 1M | Context |
|---|---|---|---|---|
| Claude Opus 5.5 | $4 | $0.2 | $20 | 1M |
| Claude Sonnet 5.5 | $2 | $0.1 | $10 | 1M |
| Claude Haiku 5.5 | $0.1 | $0.01 | $0.5 | 1M |
The catch is that coding agents read a lot of context. Every turn re-sends the conversation and the files in play, so a long session can run to millions of tokens — most of them cached input, which is billed at the much lower cached rate. As a rough guide, assuming 85% cached input, 12% fresh input and 3% output, $60 buys about 48M tokens of Claude Opus 5.5 traffic or 96M of Sonnet 5.5. The coding agent cost calculator does the arithmetic for your own usage.
A third option: a monthly plan in dollars
A Token Plan sits between the two. You pay a flat monthly price and get three times that in usage at list price — and because it is dollars rather than windows, there is no 5-hour or weekly limit:
| Plan | You pay / month | Usage at list price | ≈ tokens on Claude Opus 5.5 | ≈ tokens on Claude Sonnet 5.5 | ≈ tokens on Kimi K3 |
|---|---|---|---|---|---|
| Solo | $20 | $60 | 48M | 96M | 56M |
| Pro | $100 | $300 | 240M | 480M | 280M |
| Max | $200 | $600 | 480M | 960M | 560M |
Token estimates assume coding-agent traffic: 85% cached input, 12% fresh input, 3% output.
A plan key works in Claude Code the same way an API key does — two environment variables, described in the setup guide — and it also covers GPT, Gemini, GLM, Kimi and DeepSeek. Claude Fable 5.1 and Claude Fable 5 are not on plans.
Which one to pick
- Pro if you use Claude Code a few hours a week and rarely hit the limit.
- Max or Ultra if you only use Claude, use it daily, and can live with windows — Anthropic doesn't publish what a window is worth, but heavy Claude-only users may get the most Claude this way. Max and Ultra also include Fable.
- An API key if your usage is irregular and you want to pay only for what you use.
- A Token Plan if you want a fixed monthly bill, no windows, and the freedom to switch models — Opus for the hard parts, Kimi or GLM for the bulk.
Prices on this page come from the model catalog and were current on November 12, 2026.