💳 Data verified 2026-08-28
Prices, quotas, models, tool support and China access, side by side.
| Platform | Entry price | Tiers | Models | Usage mechanics | Tool support | China access | Status |
|---|---|---|---|---|---|---|---|
| GLM Coding Plan Best Value Z.ai (Zhipu) | $18/mo | Lite · Pro · Max | GLM-5.3 · 5.3-Flash · 5.2 · 5.1 · 5-Turbo | Weekly credits, concurrency-limited | Claude Code · ZCode · OpenCode · OpenClaw · 20+ | RMB direct | Available |
| Claude (Anthropic) Flagship Anthropic | $20/mo | Pro · Max 5× · Max 20× | Opus 4.7 · Sonnet 4.6 · Sonnet 4.5 | Rolling 5h windows + weekly caps | Claude Code · Claude app · Cowork | Intl payment | Available |
| Codex (OpenAI) OpenAI (ChatGPT) | $20/mo | Plus · Pro 5× · Pro 20× | GPT-5.6 (Sol · Terra · Luna) | Rolling 5h windows + weekly caps | Codex CLI · Codex app · IDE · GitHub | Intl payment | Available |
| MiniMax Token Plan MiniMax | $10/mo | Starter · Plus · Max · Max HS | M3 · M2.7 · M2.7-highspeed | Prompts per 5h window | Claude Code · OpenCode · Kilo Code · more | Intl payment | Available |
| Kimi Code New Model Moonshot AI | ¥49/mo | Andante · Moderato · Allegretto · Allegro | Kimi K3 (2026-07) · K2.7 | API calls per 5h, weekly refresh | Claude Code · Kimi CLI · more | RMB direct | Sign-ups paused |
| Qwen Token Plan Alibaba Cloud (Bailian) | ¥39/mo | Early bird | Qwen3.8-Max · DeepSeek-V4-Pro · qwen3.7-plus · qwen3-coder | Credits per 5h + weekly pool | Qwen Code · Claude Code-compatible endpoint | RMB direct | Available |
| OpenCode Go Lowest Entry OpenCode | $5 first mo | Go | Grok 4.5 · GLM-5.2 · Kimi K3 · DeepSeek V4 · MiniMax M3 · 18 models | Usage-value pools (5h / week / month) | OpenAI / Anthropic-compatible APIs, any tool | Intl payment | Available |
| Ark Coding Plan Hot Deal Volcengine (ByteDance) | ¥40/mo | Lite · Pro | Doubao-seed-2.0-pro | Calendar-month billing cycle | Claude Code-compatible endpoint · more | RMB direct | Available |
Prices are the latest official list prices; intro offers and yearly discounts are listed in the detail cards below.
GLM Lite at $18/mo ($12.6 yearly), or Ark Lite at ¥40 (intro ¥9.9).
Claude Max 5× ($100/mo) — the Opus family leads long-horizon code work.
Codex (Plus $20 up) — PR/issue delegation and automated code review.
GLM, Ark, Kimi and Qwen plans take RMB payment with direct access.
Skip subscriptions — pay-as-you-go API is cheaper under the usage breakpoint.
One entry subscription + pay-as-you-go API on the side; upgrade only in heavy months.
The tables above are the short version. This section covers what a coding plan actually is, how this comparison is maintained, and where to go next.
A coding plan is a fixed monthly subscription that unlocks a pool of AI coding-model usage inside agent tools — Claude Code, Codex CLI, OpenCode, ZCode and the like — instead of paying per token on an API. You pay $5–$200 (or ¥9.9–¥699) a month, and in exchange you get a quota that refreshes by the hour, day, week or month depending on the provider.
The category barely existed before 2026. GLM's Coding Plan popularized the model in the Chinese developer community, then Kimi, MiniMax, Alibaba's Qwen and Volcengine's Ark followed, while Anthropic and OpenAI folded their coding agents into Claude and ChatGPT subscriptions. Today every plan competes on three axes: price, quota mechanics, and which tools the quota works in.
The catch is always quota mechanics. A $20 plan is not "unlimited" — it is a usage pool with refresh windows (5-hour or weekly), concurrency limits and peak-hour throttling. That is why this page keeps the usage mechanics column separate from price: two plans at the same price can differ 5× in real throughput.