What does the 5-hour rolling limit mean?

Most platforms use a rolling 5-hour window to limit usage. For example, GLM Lite allows ~80 prompts per 5 hours. MiniMax has moved to monthly token-based quotas (Plus: 6B+ tokens/month). Your quota automatically refreshes as time passes.

Refund policies vary by platform. Most do not offer refunds after subscription. We recommend starting with a lower tier to test before upgrading.

2026 · Updated regularly

AI Coding Plans
Compared Side by Side

Q: What is a Coding Plan?

A Coding Plan is a monthly subscription that gives you unlimited or high-volume access to AI models inside coding tools like Claude Code, Cursor, Cline, and more. Instead of paying per token, you pay a flat monthly fee and code with AI assistance all month long. Note: Alibaba Cloud Lite has been discontinued and Coding Plans are becoming scarce.

Q: Which plan offers the best value?

For budget-conscious developers, GLM Lite at $3/month or MiniMax Plus at $20/month are excellent entry points. For heavy usage, MiniMax Max or GLM Pro offer strong value. Claude Code Pro at $20/month remains the gold standard for model quality with Fable 5.

Claude Code · GLM · MiniMax · Kimi · Qwen
Pricing, models, usage limits, and supported tools — find the plan that fits your workflow

Platforms

Lowest / mo

10+

AI Models

20+

Dev Tools

Quick Comparison

Entry-level plans at a glance — find your starting point

Scroll horizontally for full table

Platform	Starting Price	Core Models	Usage (5h window)	Tiers	Highlight
Claude Code	$20/mo	Fable 5, Opus 4.8, Sonnet 4.6	Usage limits apply	3 tiers	Gold Standard
GLM (Z.ai)	$3/mo	GLM-5.2, GLM-5.1, GLM-5, GLM-4.7	~80 prompts	3 tiers	Free MCP Tools
MiniMax	$10/mo	M3, M2.7, M2.7-highspeed	100 prompts	6 tiers	100+ TPS
Kimi	~$7/wk	Kimi K2.6	300–1,200 calls	Membership	100 TPS
Qwen (Alibaba)	$50/mo	Qwen3.6-Plus, GLM-5, Kimi	6,000 requests	1 tier	Limited Restock

Detailed Plans

Deep dive into each platform's pricing and features

Claude Code Gold Standard

Anthropic's flagship AI coding agent — Fable 5 programming powerhouse, terminal, IDE, web, and desktop

Models: Fable 5 · Opus 4.8 · Sonnet 4.6

ProEntry

Includes Claude Code & Cowork Unlimited projects Memory across conversations

$20/mo

$17/mo billed annually

Max 5xPopular

5× Pro usage Higher output limits Priority access

$100/mo

Max 20xPower User

20× Pro usage Early access to features Priority at peak times

$200/mo

Available On

Terminal CLIVS CodeJetBrainsWebDesktop AppSlack

Fable 5 Extended Thinking Web Search Connectors / MCP

View Plans

GLM Coding Plan GLM-5.2

By Zhipu AI (Z.ai) — GLM-5.2 fully available since June 13, 1M context, top-tier on Code V3

Models: GLM-5.2 · GLM-5.1 · GLM-5 · GLM-4.7

LiteBudget

~80 prompts / 5 hours ~400 prompts / week Vision + Web Search MCP included

$3/mo

ProBest Value

~400 prompts / 5 hours ~2,000 prompts / week For high-frequency complex projects

$15/mo

MaxGLM-5 Access

~1,600 prompts / 5 hours ~8,000 prompts / week GLM-5.2 support · Priority access

$49/mo

est. pricing

Supported Tools

Claude CodeRoo CodeKilo CodeClineOpenCodeCrushGooseCursor20+

Vision MCP Web Search MCP Web Reader MCP GLM-5.2 SOTA

MiniMax Token Plan M3 Model

M3 flagship — frontier coding, 1M context, native multimodal, 2 months free on annual plans

Models: M3 · M2.7 · M2.7-highspeed

Standard Plans (M2.7 ~50 TPS)

StarterEntry

100 prompts / 5 hours Lightweight workloads

$10/mo

$100/yr (save $20)

Plus

300 prompts / 5 hours 3× Starter usage

$20/mo

$200/yr (save $40)

Max

1,000 prompts / 5 hours 10× Starter usage

$50/mo

$500/yr (save $100)

High-Speed Plans (M2.7-highspeed ~100 TPS)

Plus HS

300 prompts / 5 hours ~100 TPS high-speed inference

$40/mo

$400/yr (save $80)

Max HSTop Pick

1,000 prompts / 5 hours ~100 TPS high-speed inference

$80/mo

$800/yr (save $160)

Ultra HSFastest

2,000 prompts / 5 hours 20× Starter usage

$150/mo

$1,500/yr (save $300)

Supported Tools

Claude CodeRoo CodeKilo CodeClineCodex CLIOpenCodeCursorTRAEGrok CLI

Image Understanding MCP Web Search MCP 100+ TPS Referral 10% Back

Kimi Code Plan K2.6

By Moonshot AI — Kimi K2.6 model, long-range coding + Agent cluster, weekly quota refresh

Model: Kimi K2.6

Kimi CodeMembership Perk

300–1,200 API calls per 5-hour window Up to 100 tokens/s output speed Max concurrency: 30 Weekly quota refresh cycle

~$7/wk

Part of Kimi membership

Supported Tools

Kimi CLIClaude CodeRoo Code

Kimi K2.6 Agent Cluster One-Click Auth Weekly Refresh

Get Started

Qwen Coding Plan Limited Restock

By Alibaba Cloud — Lite discontinued, only Pro remains with daily limited restock at 9:30 AM

Models: Qwen3.6-Plus · Qwen3-Max
Qwen3-Coder-Plus · GLM-5
Kimi-K2.5 · MiniMax-M2.5

LiteDiscontinued

1,200 requests / 5 hours 18,000 requests / month New purchases stopped Mar 20

$10/mo

Discontinued

ProOnly Tier Left

6,000 requests / 5 hours 45,000 requests / week 90,000 requests / month

$50/mo

Daily limited restock at 9:30 AM

⚠️ Lite plan discontinued · Pro restocked daily in limited quantities, usually sells out within minutes

Quota Reset Rules

5h rolling refreshWeekly resetMonthly reset on billing date

Qwen3.6 Series Third-Party Models Limited Restock

Frequently Asked Questions

Common questions about AI Coding Plans

What is a Coding Plan?

A Coding Plan is a monthly subscription offered by various AI platforms that gives you access to their models inside coding tools like Claude Code, Cursor, Cline, and more. Instead of paying per token, you pay a flat fee and get a generous usage quota that refreshes on a rolling basis. Note: Alibaba Cloud Lite has been discontinued, and Coding Plans are becoming increasingly scarce.

Which coding tools are supported?

Most plans support Claude Code, Cursor, Cline, Roo Code, Kilo Code, and OpenCode. GLM and MiniMax support 10–20+ tools. Kimi works primarily through Kimi CLI, Claude Code, and Roo Code. Claude Code's own plan works across terminal, VS Code, JetBrains, web, and desktop.

What does "5-hour rolling limit" mean?

Most platforms use a rolling 5-hour window to manage usage. For example, GLM Lite allows ~80 prompts per 5-hour window. MiniMax has moved to monthly token-based quotas (Plus: 6B+ tokens/month). As time passes, your oldest usage "falls off" and quota frees up automatically.

How many model calls does one prompt use?

In coding tools, a single user prompt typically triggers 5–30 model calls behind the scenes, as the tool performs code analysis, search, generation, and validation steps. Different platforms count usage differently — some count "prompts" (user messages), others count raw API requests. Check each platform's docs for specifics.

Which plan offers the best value?

It depends on your needs. For the lowest entry price, GLM Lite at $3/month is unbeatable. For a balance of price and performance, MiniMax Plus at $20/month gives you 6B+ tokens with the new M3 model. For the best model quality, Claude Code Pro at $20/month gives you access to Fable 5. For multi-model flexibility, Qwen's plan lets you switch between 6+ models (Pro only, limited restock).

Can I get a refund?

Refund policies vary by platform. Most subscriptions are non-refundable once activated. We recommend starting with a lower tier to test the service before committing to a higher plan.

AI Coding PlansCompared Side by Side

Quick Comparison

Detailed Plans

Claude Code Gold Standard

GLM Coding Plan GLM-5.2

MiniMax Token Plan M3 Model

Kimi Code Plan K2.6

Qwen Coding Plan Limited Restock

Frequently Asked Questions

Got your Coding Plan?Code on mobile with kkcode.app

AI Coding Plans
Compared Side by Side

Got your Coding Plan?
Code on mobile with kkcode.app