Claude pricing splits into two separate systems, and most confusion comes from mixing them up.
- Plans pay for the Claude apps: claude.ai, Claude Desktop, the mobile apps and Claude Code. You pay a flat monthly price and get a usage limit.
- The API pays for tokens. Every input and output token has a listed price, and you pay for exactly what you send and receive.
I list both below, as Anthropic published them on 09/28/2026, with the source for each number. Prices change. When one of them does, I update this page and the date at the top.
Claude plan prices
These are the prices on claude.com/pricing, read 09/28/2026. They exclude tax.
| Plan | Monthly price | Billed annually | Usage, as Anthropic states it | Claude Code |
|---|---|---|---|---|
| Free | $0 | $0 | Baseline | Not included |
| Pro | $20 | $17 a month ($200 up front) | At least 5x Free per five-hour session | Included |
| Max 5x | $100 | Monthly only | 5x Pro per five-hour session | Included |
| Max 20x | $200 | Monthly only | 20x Pro per five-hour session | Included |
| Team Standard | $25 a seat | $20 a seat | More usage than Pro | See pricing page |
| Team Premium | $125 a seat | $100 a seat | 5x Team Standard seats | See pricing page |
| Enterprise | Contact sales | $20 a seat, plus usage | “Usage cost scales with model and task” | See pricing page |
Two things in that table matter more than the prices.
First, no plan lists a token count. Anthropic describes each plan’s usage relative to another plan, never in tokens or messages. That is why nobody can honestly tell you that Pro “equals” some dollar amount of API usage. I cover what the limits actually are in Claude usage limits, explained.
Second, every app counts against the same limit. Anthropic’s support article on usage limits says your usage on claude.ai, Claude Code and Claude Desktop “counts towards the same usage limit.” A long Claude Code session and a long chat draw from one pool.
If you are choosing between the two paid individual plans, Claude Pro vs Max walks through who needs which.
Claude API prices
These are the per-token prices on Anthropic’s pricing documentation, read 09/28/2026. All prices are in US dollars per million tokens (MTok).
| Model | Input | Output | Cache write (5 min) | Cache write (1 hour) | Cache read |
|---|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $12.50 | $20 | $0.25 |
| Opus 5.5 | $4 | $20 | $5 | $8 | $0.20 |
| Opus 5, 4.8, 4.7, 4.6, 4.5 | $5 | $25 | $6.25 | $10 | $0.50 |
| Sonnet 5.5 and Sonnet 5 | $2 | $10 | $2.50 | $4 | $0.20 |
| Haiku 4.5 | $1 | $5 | $1.25 | $2 | $0.10 |
A few rules on the same page change what you actually pay:
- Sonnet 5 stays at $2 / $10. Anthropic had announced a rise to $3 / $15. The pricing page now states the launch price is the standard price and the rise will not happen.
- Cache writes cost more, cache reads cost much less. A five-minute cache write costs 1.25x the input price and a one-hour write costs 2x. A cache read costs 0.1x on most models, 0.05x on Opus 5.5 and 0.025x on Fable 5.1. The prompt caching cost breakdown works through when that pays.
- The Batch API costs half. Batch requests get 50% off input and output, and the batch discount stacks with caching.
- Long context costs the same. On Claude 4.6 and later models, the full one-million-token context window is billed at the standard price.
- Data residency costs 10% more. Pinning inference to the US (
inference_geo: "us") multiplies the price by 1.1. - Fast mode costs more. Fast mode on Opus 5.5 lists at $8 input and $40 output. It works only on Anthropic’s own API and not with batch.
Tools have their own prices. Web search costs $10 per 1,000 searches. Web fetch costs nothing beyond its tokens. Code execution includes 1,550 free hours a month, then costs $0.05 an hour.
What a token is, and why the same text costs more on new models
Anthropic’s rule of thumb is that one token is about four characters of English, or about three quarters of a word. The tokenizer used by Claude 4.7 and later models produces about 30% more tokens for the same text than the older one. So the same prompt sent to Opus 5.5 is counted as more tokens than it was on Opus 4.6, before any price difference.
That is the reason I built the Claude token counter. Paste a real prompt, pick a model, and it prices the input and output at the rates above, with the tokenizer difference applied. It is free and runs in the browser.
Claude Code pricing
Claude Code has no separate price. You reach it one of two ways:
- Through a Pro or Max plan. Claude Code draws from the same usage limit as the chat apps. When you hit the limit, you wait for the reset, upgrade, or buy usage credits. Anthropic’s Claude Code support article adds a fourth option: switch to API credits, billed at standard API rates, which Claude Code does only with your explicit consent.
- Through API credits. Every token Claude Code sends is billed at the API prices above. A long agentic session re-reads a large context on every turn, so cached input makes up most of the bill. Opus 5.5 vs Fable 5.1 walks through what a full day of agentic work costs at list prices.
Which is cheaper depends on your volume, and Claude Code usage limits covers how to make a plan last before you pay for either.
Which one to pay for
- You chat and write, a few hours a week: Free or Pro. The API costs more setup than it saves.
- You run Claude Code most working days: a plan. Pro first. Move to Max when you hit the five-hour limit often enough that waiting costs you more than $80 a month in lost time.
- You are building a product on Claude: the API. Plans pay for your own use of the Claude apps. The API is what Anthropic sells for building on Claude.
For the API path, the number that matters is not the price per million tokens. It is your tokens per task, times your tasks per month. The price table gives the first half. Your own prompts give the second.
Anthropic sells you a limit and never states its size. The person who prices their own day knows what they are buying.
Watching the prices
These numbers move. Two recent examples: Anthropic cancelled a planned Sonnet 5 price rise, and Opus 5.5 launched on 09/22/2026 at 20% under the price of Opus 5. If you budget against these numbers, a change you miss costs you money silently.
I am building Claude Cost Watch, a short email that tells you when a Claude price or limit you depend on changes. It is a waitlist today. Nothing is charged.
Sources
All read on 09/28/2026.
- Plans: claude.com/pricing
- API prices, cache multipliers, batch, fast mode, tools: platform.claude.com/docs/en/about-claude/pricing
- Shared usage limit across apps: Anthropic Help Center, “How do usage and length limits work?”
- Claude Code on a plan: Anthropic Help Center, “Use Claude Code with your Pro or Max plan”
No link on this page is an affiliate link. Anthropic has no affiliate program I could verify.