What Is a Coding Plan?
Last updated October 15, 2026.
A coding plan is a flat monthly subscription that lets a coding agent call a model through an API key. You pay a fixed fee. Usage is capped by rolling windows, usually 5 hours and 7 days, instead of billed per token. The plans exist because agent loops resend the same repo context hundreds of times a day, and per-token bills for that traffic swing from week to week.
The open-weight labs sell most of them. Z.ai sells the GLM Coding Plan. Moonshot bundles Kimi Code into Kimi membership. MiniMax sells the M Plan. Alibaba Cloud Model Studio sells a Coding Plan that covers Qwen, GLM, Kimi, and MiniMax models behind one key. Z.ai and MiniMax both list Claude Code, Codex, and OpenCode as supported tools.
Coding Plan Prices Compared (October 2026)
Monthly list prices from each vendor's subscription page, checked October 15, 2026. Annual discounts are noted where the vendor shows them.
| Plan | Price | Usage limit | Models |
|---|---|---|---|
| GLM Coding Plan Lite | $18/mo ($12.60/mo yearly) | 2,000 credits per 5h, 10,000 per week | GLM-5.3, GLM-5.3-Flash |
| GLM Coding Plan Pro | $80/mo ($56/mo yearly) | 6x Lite (12,000 per 5h, 60,000 per week) | GLM-5.3, GLM-5.3-Flash |
| GLM Coding Plan Max | $168/mo ($117.60/mo yearly) | 14x Lite (28,000 per 5h, 140,000 per week) | GLM-5.3, GLM-5.3-Flash |
| Alibaba Coding Plan Pro | $50/mo | 6,000 requests per 5h, 45,000 per week, 90,000 per month | Qwen 3.7/3.6/3.5 Plus, Kimi K2.5, GLM-5, MiniMax M2.5 |
| Kimi Plus (includes Kimi Code) | $19/mo ($15/mo yearly) | 1x agent credits | Kimi K3 |
| Kimi Pro | $39/mo ($31/mo yearly) | 2x agent credits | Kimi K3 |
| MiniMax M Plan Go | $220/yr billed annually | Base usage | MiniMax M3.1 Flash Preview |
| MiniMax M Plan Explore | $550/yr billed annually | 3x Go | MiniMax M3.1 Flash Preview |
| MiniMax M Plan Build | $1,320/yr billed annually | 7.5x Go, priority at peak hours | MiniMax M3.1 Flash Preview |
For reference, Claude Pro is $20 a month and Claude Max is $100 (5x Pro usage) or $200 (20x). Those plans cover Claude Code with Anthropic models only. The plans above put open-weight models behind the same agents for less.
GLM Coding Plan (Z.ai)
The GLM Coding Plan is the largest plan by search demand. It runs GLM-5.3 and GLM-5.3-Flash with a 1M context window, and Z.ai lists more than 20 supported tools, including Claude Code, Codex, OpenCode, Cursor, and its own ZCode IDE. All tiers include vision, the Web Search MCP, the Web Reader MCP, and Zread MCP.
Usage is metered in credits on two windows. The 5-hour credits refresh 5 hours after they are consumed. The weekly credits reset every 7 days from the subscription date. Z.ai publishes token estimates per tier at a 95% cache hit rate.
| Tier | 5-hour credits | Weekly credits | GLM-5.3 tokens/week | GLM-5.3-Flash tokens/week |
|---|---|---|---|---|
| Lite ($18) | 2,000 | 10,000 | 48M-97M | 146M-292M |
| Pro ($80) | 12,000 | 60,000 | 290M-580M | 877M-1,755M |
| Max ($168) | 28,000 | 140,000 | 676M-1,352M | 2,047M-4,095M |
For the per-token GLM API and its Claude Code setup, see the GLM 5 family guide.
Qwen and Alibaba Coding Plan
Alibaba Cloud Model Studio sells one tier to new subscribers. Coding Plan Pro costs $50 a month. It caps you at 6,000 requests per 5 hours, 45,000 per week, and 90,000 per month. All three caps apply at once, and calls pause at whichever comes first.
It is the only plan here that mixes labs. The supported models are qwen3.7-plus, qwen3.6-plus, kimi-k2.5, glm-5, MiniMax-M2.5, and qwen3.5-plus. Alibaba counts quota in model calls, not tasks. Its docs say a simple task uses 5 to 10 calls and a complex one 10 to 30 or more, so 90,000 requests is roughly 3,000 to 9,000 agent tasks a month.
The Lite plan stopped taking new subscribers on March 20, 2026, and renewals ended April 13, 2026. Pro slots are limited and restock daily at 00:00 UTC+8.
Kimi Coding Plan (Kimi Code)
Moonshot has no standalone coding plan. Kimi Code, its agent for desktop, CLI, and IDE plugins, ships with every paid Kimi membership and runs Kimi K3. Plus is $19 a month, Pro $39, Max $99, and Ultra $199. Billed annually they drop to $15, $31, $79, and $159. Pro, Max, and Ultra carry 2x, 5x, and 10x the agent credits of Plus. Max and Ultra add up to 1M tokens of context.
Moonshot does not publish credit-to-token figures on the pricing page, so the allowance per dollar is harder to compare than Z.ai's. For per-token Kimi K3 access, see the Kimi K3 API guide.
MiniMax Coding Plan (M Plan)
MiniMax renamed its coding plan the M Plan. Billed annually, Go is $220 a year, Explore $550 with 3x Go usage, and Build $1,320 with 7.5x Go usage plus priority at peak hours. Every tier includes MiniMax Code on web, desktop, terminal, and mobile, and an API key that works in Claude Code, Codex, Cursor, and OpenCode.
The current text model is MiniMax M3.1 Flash Preview. MiniMax has posted a retirement notice for M2.5 and M2.7, so configs pinned to those ids need updating.
DeepSeek Coding Plan
DeepSeek does not sell a coding plan. Its API pricing page lists per-token rates only. People searching for one usually want DeepSeek models in Claude Code at a predictable cost, which means per-token billing with caching. Morph serves DeepSeek V4.1 Flash as morph-dsv41flash at $0.12/M input and $0.8/M output on an OpenAI-compatible API.
Coding Plan vs API: Break-Even
Z.ai publishes enough data to price the GLM plan against its own API. GLM-5.3 costs $1.40/M input, $0.26/M cached input, and $4.40/M output on the Z.ai API. At a 95% cache hit rate, blended input costs 0.95 x $0.26 + 0.05 x $1.40 = $0.317 per million tokens. Multiply the weekly token estimates by 52/12 to get a monthly figure.
| Tier | Plan price | GLM-5.3 tokens/month | Same input on the API |
|---|---|---|---|
| Lite | $18 | 208M-420M | $66-$133 |
| Pro | $80 | 1,257M-2,513M | $398-$797 |
| Max | $168 | 2,929M-5,859M | $929-$1,857 |
Output tokens would add to the API column, so these figures understate the plan's value. They also assume you use the whole allowance. A developer who codes two hours a day uses a fraction of it. Bursty work hits the 5-hour cap and waits, while the API keeps answering.
The rule of thumb: if you code daily inside one agent, a plan is cheaper. If you run batch jobs, CI agents, or many parallel sessions, pay per token. For per-token numbers across Claude, OpenAI, Gemini, and DeepSeek, see AI coding costs.
Pros and Cons of a Coding Plan
- Fixed monthly cost for steady daily coding
- Allowances worth several times the fee at API rates when fully used
- One key works in Claude Code, Codex, OpenCode, and Cursor
- Vendor MCP tools such as web search are often bundled
- 5-hour caps pause calls mid-task when hit
- Plans are restricted to supported tools, not general API use
- Credit systems differ per vendor, so allowances are hard to compare
- Models and tiers change often: Alibaba closed Lite, MiniMax is retiring M2.5 and M2.7
Which Coding Plan to Pick
Lowest monthly price
GLM Coding Plan Lite at $18/month, or $12.60/month billed yearly. Z.ai publishes token estimates, so you can check the allowance against your usage before paying.
Several labs under one key
Alibaba Coding Plan Pro at $50/month covers Qwen, GLM, Kimi, and MiniMax models, capped at 90,000 requests a month. Slots are limited.
Already use Kimi
Kimi Code comes with any paid Kimi membership from $19/month, so a separate coding subscription is unnecessary.
Batch, CI, or parallel agents
Skip the plan and pay per token. Rolling 5-hour caps stall unattended jobs, and the API has no window to wait out.
Frequently Asked Questions
What is a coding plan?
A coding plan is a flat monthly subscription that gives AI coding tools such as Claude Code, Codex, and OpenCode access to a model through an API key. Instead of paying per token, you pay a fixed fee and get usage capped by rolling windows, usually a 5-hour limit plus a weekly limit. Z.ai, Moonshot (Kimi), MiniMax, and Alibaba Cloud sell them.
How much does the GLM Coding Plan cost?
As of October 15, 2026, the GLM Coding Plan is $18/month for Lite, $80/month for Pro, and $168/month for Max. Quarterly billing takes 20% off and yearly billing 30% off, which brings Lite to $12.60/month. Pro has 6x Lite usage and Max 14x. Lite gets 2,000 credits per 5 hours and 10,000 per week.
Is there a Kimi coding plan?
Moonshot does not sell a separate coding-only plan. Kimi Code, its coding agent for desktop, CLI, and IDEs, is included in every paid Kimi membership. Plus costs $19/month ($15/month billed annually), Pro $39, Max $99, and Ultra $199. Higher tiers carry 2x, 5x, and 10x the agent credits of Plus.
What is the MiniMax coding plan called now?
MiniMax renamed it the M Plan. Billed annually, Go costs $220/year, Explore $550/year with 3x Go usage, and Build $1,320/year with 7.5x Go usage. Every tier includes MiniMax Code and an API key that works in Claude Code, Codex, Cursor, and OpenCode.
What does the Alibaba (Qwen) Coding Plan include?
Alibaba Cloud Model Studio's Coding Plan Pro costs $50/month and allows up to 6,000 requests per 5 hours, 45,000 per week, and 90,000 per month. It covers qwen3.7-plus, qwen3.6-plus, kimi-k2.5, glm-5, MiniMax-M2.5, and qwen3.5-plus. The Lite plan closed to new subscribers on March 20, 2026, and slots for Pro are limited and restocked daily.
Does DeepSeek have a coding plan?
No. As of October 15, 2026, DeepSeek's API pricing page lists per-token rates only, with no subscription or coding plan. DeepSeek models are billed per token, through DeepSeek or another provider.
Is a coding plan cheaper than the API?
For steady daily coding, yes. Z.ai estimates GLM Coding Plan Lite at 48M to 97M GLM-5.3 tokens a week at a 95% cache hit rate. At Z.ai's API rates ($1.40/M input, $0.26/M cached), that input alone would cost about $66 to $133 a month, against an $18 fee. Plans lose for bursty or batch workloads, since the 5-hour cap pauses calls once it is hit and unused quota does not carry over.
The fastest endpoints are private deployments
Morph's top speeds come from dedicated deployments, not shared public endpoints: speculators trained on your traffic, caching tuned to your workload, and volume discounts over public per-token rates. Over 500 billion tokens per day run this way.
Run GLM-5.3 per token on Morph
Morph serves GLM-5.3, Kimi K3, and DeepSeek V4.1 Flash on an OpenAI-compatible API. No 5-hour window, pay for the tokens you use.
Sources
- Z.ai: GLM Coding Plan pricing (checked October 15, 2026)
- Z.ai docs: GLM Coding Plan overview, credits and token estimates
- Z.ai docs: API pricing (GLM-5.3 $1.40/$0.26/$4.40)
- Alibaba Cloud Model Studio: Coding Plan overview (updated October 11, 2026)
- Kimi: membership pricing
- MiniMax: M Plan pricing
- DeepSeek API docs: models and pricing