This is a complete comparison guide to China's 7 major Coding Plans (Volcengine/Alibaba/Tencent/Baidu/Kimi/Zhipu/MiniMax), listing each platform's models, pricing, quotas, and characteristics. This essentially represents the full landscape of the domestic AI Coding ecosystem.

Coding Plans are subscription packages for AI programming models offered by cloud providers.

Core Characteristics:

  • Fixed monthly fee
  • Provides large AI request quotas
  • Can be used in tools like Cursor, Claude Code, Cline, OpenClaw, etc.

Essentially: Coding Plan = AI programming model subscription service.

1. Volcengine Ark Coding Plan

Platform: Volcengine

Supported Models:

  • Doubao-Seed-Code
  • DeepSeek-V3.2
  • GLM-4.7
  • Kimi-K2

Users can freely switch between models or use Auto mode for automatic selection.

Subscription Tiers:

TierPriceRequest Quota
LiteFirst month ~9.9 yuan, 40 yuan/month thereafter1,200 times per 5 hours
ProFirst month ~49.9 yuan, 200 yuan/month thereafter6,000 times per 5 hours

Characteristics:

  • Multi-model aggregation
  • Auto automatic model selection
  • Compatible with Cursor / Claude Code / Cline
  • Primarily focused on ByteDance ecosystem models

2. Alibaba Cloud Coding Plan

Platform: Alibaba Cloud

Supported Models:

  • Qwen3.5-Plus
  • Qwen3-Coder
  • GLM-5
  • MiniMax-M2.5
  • Kimi-K2.5

Users can freely switch between multiple models.

Subscription Tiers:

TierPriceRequest Quota
Lite40 yuan/month18,000 times per month
Pro200 yuan/month90,000 times per month

Characteristics:

  • Qwen programming models are very strong
  • Large number of models available
  • Strong Agent ecosystem
  • Has usage restrictions (only allowed for AI Coding tool calls)

3. Tencent Cloud Coding Plan

Platform: Tencent Cloud

Supported Models:

  • Tencent HY 2.0
  • GLM-5
  • Kimi-K2.5
  • MiniMax-M2.5

Subscription Tiers:

TierPrice
LiteFirst month 7.9 yuan
ProFirst month 39.9 yuan

Renewal Pricing:

  • Lite: 40 yuan/month
  • Pro: 200 yuan/month

Characteristics:

Advantages:

  • Tencent Hunyuan models
  • Strong enterprise ecosystem

Disadvantages:

  • Programming model maturity slightly weaker than Qwen

4. Baidu Qianfan Coding Plan

Platform: Baidu

Supported Models:

  • GLM-5
  • MiniMax-M2.5
  • Kimi-K2.5
  • Wenxin series models

Characteristics:

  • Qianfan ecosystem
  • Many enterprise users
  • Smaller developer community

5. Kimi Code

Platform: Moonshot AI

Core Models:

  • Kimi K2
  • Kimi K2.5

Characteristics:

  • Long context capability
  • Strong reasoning ability

Suitable For:

  • Complex project analysis
  • AI Agent applications

6. Zhipu GLM Coding Plan

Platform: Zhipu AI

Supported Models:

  • GLM-4.7
  • GLM-5

Characteristics:

  • Strong reasoning capabilities
  • Agent ecosystem

7. MiniMax Token Plan (Added 2026-03-26)

Recently been using MiniMax for programming with good results, a perfect replacement for my previous ChatGPT usage.

Supported Models:

  • MiniMax 2.7
  • Speech-2.8-HD (voice model)
  • MiniMax Hailuo 2.3 (video model)
  • image-01 (image model)
  • music-2.5+ (music model)

Characteristics:

  • M2.7 programming flagship
  • Ultimate cost-performance ratio
  • All-modal one-stop solution

Pricing:

Screenshot shows monthly subscription pricing; annual packages are cheaper. Lowest is 29 yuan/month—pay attention to call limits.

Six Major Coding Plan Comparison

VendorPriceCharacteristics
Alibaba Cloud40 yuan/monthQwen strongest, multi-model switching
Tencent Cloud40 yuan/monthHunyuan ecosystem, multi-model switching
Baidu Cloud39.9 yuan/monthEnterprise ecosystem, multi-model switching
Kimi49 yuan/monthLong context capability
Zhipu49 yuan/monthStrong reasoning, SOTA level
MiniMax29 yuan/monthUltimate cost-performance, SOTA level

Programming Capability Rankings

For programming capability rankings, refer to SWE-bench, where GLM 5, MiniMax 2.5, Kimi 2.5, and DeepSeek 3.2 are all on the leaderboard.

Personal Experience

Previously always used ChatGPT for programming, from ChatGPT 5.0 through ChatGPT 5.3.

Currently switched entirely to domestic large models, mainly using GLM and MiniMax—can say they perfectly replaced ChatGPT.

Additional note: DeepSeek doesn't have particular standout features currently; will supplement later.

Selection Recommendations

For Budget-Conscious Developers:

  • MiniMax offers the best value at 29 yuan/month with SOTA-level performance

For Long Context Needs:

  • Kimi Code specializes in long-context processing for complex project analysis

For Multi-Model Flexibility:

  • Alibaba Cloud and Volcengine offer the most model variety with easy switching

For Enterprise Integration:

  • Tencent Cloud and Baidu Qianfan have strong enterprise ecosystems

For Raw Programming Performance:

  • Alibaba Cloud's Qwen models currently lead in coding benchmarks

Important Considerations

  1. Usage Restrictions: Some platforms restrict API usage to AI Coding tools only
  2. Quota Management: Understand whether quotas are per-hour, per-day, or per-month
  3. Model Updates: The AI landscape evolves rapidly—verify current model availability
  4. Integration Compatibility: Ensure your preferred IDE/tools support the platform

This guide reflects the market state as of early 2026. Pricing, models, and features subject to change. Always verify current offerings on official platforms before subscribing.