Price-per-token arbitrage across 37 Chinese AI platforms

What it does codingplan aggregates subscription plans and API pricing from 37 Chinese AI platforms—Zhipu, Kimi, MiniMax, ByteDance Fangzhou, DeepSeek, and others—into sortable comparison tables. It tracks which models are included in each tier and normalizes disparate billing schemes into a single figure: RMB per million tokens.

The interesting bit The README is auto-generated from a dataset mapping 1,090 package-model relationships, which is why it feels more like a broker’s price sheet than typical repo documentation. The maintainer even labels rates as “calculated” or “measured/tested,” admitting when numbers are derived rather than verified.

Key highlights

• Covers 118 models across 99 subscription packages and pay-as-you-go APIs
• Compares peak versus off-peak pricing and subscription allowances versus API credits
• Flags availability status: open purchase, suspended, or flash-sale only
• Notes international editions (e.g., Zhipu international) and their pricing premiums
• Published as a static site with platform, usage, and plan-specific views

Caveats

• Pages are interspersed with referral links (api.dreamfree.space), so the project doubles as a curated affiliate directory
• “Calculated” and “measured” comparison methodology is labeled but not explained in detail
• It is a data reference and HTML frontend, not a deployable tool or library

Verdict Useful if you are a Chinese developer optimizing inference spend across competing AI platforms. Skip it if you want open-source infrastructure to self-host or if you work outside the Chinese AI subscription market.