Claude Code API Pricing: Top-Ups, Tiers, and Channels

Pay for dollar quota in top-ups from $50 to $2,000, valid at least 30 days, then spend it on Claude, GPT, or Gemini at your channel's rate.

Two separate numbers decide what a month of coding costs. The first is the price of quota: a top-up of $100 costs $15.99, and larger top-ups get cheaper per dollar. The second is how fast a channel consumes that quota: AWSQ takes 0.8× per request, CCMAX 2×, Codex 0.8×, and Gemini 0.5×. Nothing else in this product changes the amount of quota you bought.

Top-up prices

Top-ups are cumulative tiers, not all-or-nothing discounts. The first $100 of quota in an order is billed at $0.1599 per dollar of quota, the next $400 at $0.1475, and everything above $500 at $0.13.

Because the bands stack, an amount between the published anchors lands between them. A $250 top-up is 100 × $0.1599 plus 150 × $0.1475, and the checkout total is rounded once at the end so the displayed price and the charge agree to the cent.

The $50 floor matters for small experiments: the 100-dollar band's rate applies to the whole order, so a minimal top-up costs about $8.

The table shows standard USD reference prices. Checkout currency depends on your region and Stripe subscription status. With a Stripe subscription, the currency is fixed to SGD in the China region and USD elsewhere. Without one, you can pay in HKD, with SGD also available in the China region. Eligible legacy accounts without a Stripe subscription can also select CNY.

Channel multipliers

The multiplier is applied to each request's cost at the configured model rates. It changes how long a balance lasts, never the balance itself.

A rough mental model: $100 of quota covers about $125 of Claude usage at the configured base rates on AWSQ and about $50 on CCMAX, for the identical request. Choose CCMAX when you need Fable or a second Claude route; choose AWSQ for volume.

How an estimate is built

The calculator at the bottom of this page uses one formula, and you can reproduce it by hand:

1. Reference cost. Multiply uncached input, cache-read, and output tokens by the vendor's published per-million rates for that model, then divide by 1,000,000.
2. Quota required. Price those same counts at the current AI in IDE model rates and apply the channel multiplier. The calculator reads the configured rates used for billing.
3. Rounded top-up. Round the quota requirement up to the next $25, with a $50 floor, since that is what the slider accepts.
4. Top-up price. Run the rounded amount through the tier table above.

For Claude, select AWSQ or CCMAX and enter the number of billable requests if known. Each request consumes at least $0.0004 of quota on AWSQ or $0.001 on CCMAX. The calculator uses standard rates before membership discounts. Pro, Custom and Team members may use less quota than shown. With the same undiscounted rates, a mix of short and long requests may use more.

Standard provider reference rates as of November 2, 2026. The Gemini 3.8 Flash rates below apply through December 31, 2026.

Rates are per 1M tokens in USD. Two details shift real bills away from a clean estimate: when the upstream actually serves Opus 5.5 in Fast mode, its base and cache rates double, and cache writes cost more than cache reads — 1.25× the input rate for the default five-minute write, 2× for the one-hour write. Billing follows the usage block in the response, so a request served in Fast mode is charged as Fast even if you did not ask for it.

The calculator is interactive and sits below this section. Enter your own token counts there rather than reading a number off this page; any figure printed here would be stale the moment your input/output mix changed.

Configuring Claude Code

Three values connect a Claude Code client to this service:

Copy
1
export ANTHROPIC_BASE_URL='https://api.aiinide.com'
2
export ANTHROPIC_AUTH_TOKEN="YOUR_DASHBOARD_API_KEY"
3
export API_TIMEOUT_MS=600000

ANTHROPIC_BASE_URL must be the bare host. Claude Code appends /v1/messages; writing https://api.aiinide.com/v1 produces /v1/v1/messages and a 404. Put the same three values in ~/.claude/settings.json under env if the editor was launched from the Dock or Start menu, because those processes do not inherit your shell profile.

For OpenAI-compatible clients the base URL becomes https://api.aiinide.com/v1, and the key is sent as Authorization: Bearer .... For the claude command, nothing else changes: /status reports the active model, and Usage shows dollars and tokens per key.

Step-by-step installs live in Configuration, and editor-specific notes in Claude Code in VS Code, Cursor, and JetBrains.

FAQ

Is a top-up the same as a subscription? A subscription renews each billing cycle. A top-up is a one-time purchase valid for at least 30 days and is spent before subscription quota. When a gift key runs out, switch your client to a standard account key to use the top-up.

What happens to unused top-up quota? It expires when the top-up's window ends: 30 days after purchase, or your subscription end date if that is later, so a top-up never dies before the subscription it was bought alongside.

How does CCMAX consumption compare with AWSQ? CCMAX uses 2× the configured model rate and AWSQ uses 0.8×. The same request consumes 2.5 times as much quota on CCMAX.

Can I use several channels at the same time? Yes. Create a standard account key for each channel. They share the account balance, and each channel applies its own multiplier.

Do the tier prices apply per purchase or per account? Per purchase. Each order restarts at the $0.1599 band, so two $100 orders cost more than one $200 order.

Does the estimate include web search or other tool fees? No. The calculator covers token usage only. Server-side web search and metered tools are billed separately per call at the rates in the channel documentation. The Gemini channel rejects hosted search tools.