Claude Code Pricing [2026 Edition] — Subscription and API Pay-as-You-Go, Actual Costs and Break-Even Point

Claude Code has two completely separate pricing models, and almost all confusion about its cost comes from mixing them up. A claude.ai subscription includes the tool in a flat monthly plan with usage limits. An API key bills per token, with no monthly fee and no cap. This page brings together current figures for both options, a calculation to help you choose, the cost of a real month of work, and how to pay from Japan.

Rates verified on October 10, 2026 against claude.com/pricing and Anthropic's API pricing documentation. Kunavo's per-token rates are read live from the model catalog.

The two options side by side

What the subscription costs

Claude Code is included in claude.ai's paid plans. Anthropic's setup documentation specifies that the free plan does not include Claude Code. Pro costs $20/month ($17/month when billed annually, $200 upfront) and covers one developer working within the plan limits. Max starts at $100/month, for those who keep hitting those limits. Team and Enterprise add per-seat quotas and spending controls for administrators.

Two points matter from Japan: billing is recurring in dollars, so you need a credit card that supports international payments, and your card issuer's foreign transaction fee applies every month. The other point is that usage limits are shared with Claude chat—an afternoon spent chatting uses the same window as your coding session.

What does the API route cost?

There is no monthly fee; you pay per token. Prices are in USD per 1M tokens, with Anthropic's official rates alongside for a direct comparison.

Cached input is billed at 10% of the input rate, and a five-minute cache write at 125%. Because Claude Code rereads the entire conversation at every step, these two multipliers drive the real-world bill more than model choice does.

The real cost of one step

An “agentic step” is one back-and-forth with the model within a task: system prompt + conversation + file context in, one edit or explanation out. Actual measurements are approximately 25,000 input / 1,200 output tokens. At Kunavo rates:

Break-even point

The break-even point depends on the number of steps, not the number of workdays. If Sonnet 5 costs $0.043 per step, a $20/month Pro subscription is equivalent to about 460 steps, or around 23 tasks of 20 steps each. Below that, pay-as-you-go is cheaper, and months without use cost $0. Above it, the subscription is better until you hit the plan limits.

To calculate using your own actual figures, run /usage in Claude Code, obtain the four token counts, and enter them as they are:

# Claude Code 1 セッションの実費を、そのセッションのトークン数から出す。
# 4 つの実数は Claude Code の中で /usage を実行すれば取れる。
RATES = {  # Kunavo の 1M トークンあたり USD: (input, output)
    "claude-haiku-4-5":  (0.70, 3.50),
    "claude-sonnet-4-6": (2.10, 10.50),
    "claude-sonnet-5":   (1.40, 7.00),
    "claude-opus-5":     (3.50, 17.50),
}

def session_cost(model, inp, out, cache_read=0, cache_write=0):
    i, o = RATES[model]
    return (
        inp / 1e6 * i                   # 新規入力
        + out / 1e6 * o                 # 出力（thinking トークンを含む）
        + cache_read / 1e6 * i * 0.10   # キャッシュ読み出し: 入力レートの 10%
        + cache_write / 1e6 * i * 1.25  # 5 分キャッシュ書き込み: 125%
    )

print(session_cost("claude-sonnet-5", 1_200, 5_300, 940_000, 50_000))

Point Claude Code to Kunavo

Claude Code works with any Anthropic-compatible endpoint. Set two environment variables to switch to pay-as-you-go without changing your CLI installation or workflow:

export ANTHROPIC_BASE_URL=https://api.kunavo.com
export ANTHROPIC_AUTH_TOKEN=sk-kn-...
export ANTHROPIC_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-5-5
export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5

The model specification line prevents the first request from returning 404 when Kunavo does not yet offer the latest Opus targeted by Claude Code’s default model and the opus alias. Because the sonnet alias points to Sonnet 5.5, which Kunavo does not offer, without the ANTHROPIC_DEFAULT_SONNET_MODEL line, /model sonnet, the execution phase of opusplan, and subagents configured with model: sonnet will all return 404. The opus alias is fixed to Opus 5.5 (claude-opus-5-5), and requires Claude Code v2.1.280 or later. If your version is older, update it with claude update. Issue keys at /app/keys. For detailed steps, see the guide to configuring an API key for Claude Code; if you encounter an error, see How to resolve a 401 error.

For business use

Team / Enterprise subscriptions add per-seat billing and spending controls for administrators, with a monthly charge for each seat. The API key route has no concept of seats: multiple people and machines can share a single prepaid balance, while separate keys let you track usage by purpose (per developer, CI, or batch processing). In months when not all seats are used, or when much of the automated usage is not tied to a person, pay-as-you-go makes costs easier to track than per-seat billing. Usage can be retrieved by key through the Usage API.

Four ways to pay less

1. Let prompt caching work. Cached input costs 10% of the input rate. Claude Code rereads the whole conversation at every step, making this the single most effective lever. See the cache documentation.
2. Clear context between unrelated tasks. Carrying one long session forward makes input tokens keep growing at every step.
3. Choose a model based on the size of the problem. Do not default to the biggest model. As the table above shows, Haiku and Opus differ by an order of magnitude per step.
4. Keep the Haiku line. Without ANTHROPIC_DEFAULT_HAIKU_MODEL, the haiku alias calls a model ID that Kunavo does not offer (Haiku 5.5 from v2.1.293 onward), causing /model haiku and subagents configured with haiku to return 404s. The same line also moves Claude Code's own auxiliary calls from the main model to a cheaper model.

Paying from Japan

The subscription is recurring billing in dollars and requires a credit card enabled for international payments. The API key route uses Stripe checkout and supports cards including JCB (Visa, Mastercard, Amex, JCB, UnionPay), Apple Pay, Google Pay, and Link. It uses a prepaid wallet model, with a minimum top-up of $10 and no balance expiration. There is no charge in months when you do not write code. Information required for a domestic business is compiled in the disclosure under the Specified Commercial Transactions Act. Billing details are in the billing documentation.

Practical details beyond pricing—how to write CLAUDE.md commands, set permissions, and keep working after usage limits close—are covered in how to use Claude Code. For a comparison with Codex's pay-as-you-go pricing, see Codex pricing, which uses the same calculation method. Options for when you hit a subscription usage limit are explained in Claude Pro limits.

Frequently asked questions