Get started

Pricing

Last updated: October 28, 2026

CoreCode's unit price for each model in each group, next to its vendor's official price. This table syncs from the console every 10 minutes; orders and charges follow the console's Pricing page.

Looking for Anthropic's own API prices rather than CoreCode's? Every model's official input, output and cache prices, the batch discount, fast mode and a worked Claude Code example are in Claude API pricing.

How billing works

• Pay per use. No subscription or plan: each request is charged for what it used, and every one is in your usage logs.
• ¥1 buys $1 of credit. You top up with Alipay or WeChat Pay; the console shows balance and costs in dollars of credit. The table below converts that into the yuan you pay.
• Four kinds of tokens, priced separately. Input, output, cache writes and cache reads, on the same basis as each vendor. Most of Claude Code's context goes through cache reads, at 5% of the input price on Opus 5.5 and Sonnet 5.5, so what you spend depends mostly on the cache hit rate; see Caching and billing.
• The 1-hour cache writes at about twice the input price; the table lists the default 5-minute cache write.

Which group to choose

You pick a group when you create a key, and the key calls only that group's models. Prices are in the table below:

• Claude 满血（仅限CC） (Claude Code only): admits the Claude Code CLI and desktop app. Choose it for everyday Claude Code.
• Claude 满血（可外接） (external clients): the same official Claude accounts, and also admits SDKs, curl and other clients.
• GPT, Gemini, Grok: one group per vendor, called through Codex or the OpenAI-compatible API.
• auto: one key for every group; the model picks the group.

Whether each Claude model is the one you chose is checked every 3 days with the open-source ModelTrace fingerprint; the results and raw data are in the check report.

The models in each group and the API each uses are in Supported models.

Prices by model

Group 「Claude 满血（仅限CC）」

CNY per million tokens; official prices at $1 = ¥6.8. 72% below the official price.

claude-haiku-5-5: the table applies up to 100,000 tokens of context; above that, long-context prices apply: input ¥0.95, output ¥4.75, cache write ¥1.1875, cache read ¥0.095.

claude-fable-5-1 costs more to serve, so it is priced at 1.5× the official price before the group discount. That is 58% below the official price.

Group 「Claude 满血（可外接）」

CNY per million tokens; official prices at $1 = ¥6.8. 70% below the official price.

claude-haiku-5-5: the table applies up to 100,000 tokens of context; above that, long-context prices apply: input ¥1.025, output ¥5.125, cache write ¥1.2812, cache read ¥0.1025.

claude-fable-5-1 costs more to serve, so it is priced at 1.5× the official price before the group discount. That is 55% below the official price.

Group 「GPT」

CNY per million tokens; official prices at $1 = ¥6.8. 88% below the official price.

Group 「Gemini」

CNY per million tokens; official prices at $1 = ¥6.8. 63% below the official price.

Group 「Grok」

CNY per million tokens; official prices at $1 = ¥6.8. 82% below the official price.

Next

Anthropic's full API price list, with the batch discount and long-context rules, is in the blog post Claude API pricing. To get the same work done for less, see Spending less on Claude Code. New here? Follow the new user guide; the newcomer pack is ¥29.9 for $50 of credit, see Top-up.

FAQ