GLM Coding Plan: Which Option Should You Choose?

Three ways to buy GLM access

The same model family can come with different billing rules, credentials and tool access. Decide which service you need before comparing the size of its allowance.

On smaller screens, scroll the tables sideways to compare all columns.

See our model catalog and Multi-Model plans. Z.ai lists official plans from $18 per month; confirm the selected billing term and any offers at checkout.

The official plan currently uses both 5-hour and weekly credits. Its calculation distinguishes input, cached input and output, with lower credit consumption off peak. Our gateway uses a points rate for each model. One official credit is not one gateway point. Comparing the headline balances alone will not tell you which option costs less.

Z.ai uses its own credit calculation and limits subscription access to its supported tools and products.

Which GLM models does our plan include?

Our current Multi-Model catalog includes GLM 5.3 and GLM 5.3 Flash. The table below describes our gateway configuration. Enter the exact gateway ID in your coding tool; a display name is not always the ID the API expects.

Source: our published model rates and the model configuration used by our gateway setup guides. Both GLM routes have reasoning enabled; Low reduces effort rather than switching reasoning off. The configured default is Max.

At these rates, the same token count uses 9 times as many points on GLM 5.3 as on Flash. For an initial evaluation, try Flash on a representative task, then compare GLM 5.3 on a task where the extra points would be justified by a better result. Check the resulting code and tests; the higher rate alone does not establish better performance on your project.

If your agent needs to inspect screenshots, select the image-capable Flash route. Our current GLM 5.3 route is text-only, regardless of capabilities available through other providers.

How much GLM usage does a plan provide?

Our Multi-Model plans meter points across two limits: a 5-hour allowance and a 7-day allowance. Every model call counts toward both. Switching to another model changes the points rate, but does not reset either limit.

Cached input is counted once within input usage. Reasoning tokens are counted once within output usage. An agent can make several model calls during one task, so count the full session rather than just the final answer.

This is a price summary, not a checkout quote. Pricing shows current stock, one-time purchase options and recurring billing terms. The figures above are points, not an equal number of tokens for every model.

Example: a mixed GLM session on the entry plan

Assume you start with both allowances unused and make all of these calls within one 5-hour window. The token totals below are hypothetical, not a benchmark or a promise about how many tasks you can complete.

If you used the entry plan’s full 60M-point 5-hour allowance exclusively on GLM 5.3, it would cover approximately 6.67M combined input and output tokens. On GLM 5.3 Flash, it would cover 60M. Both examples require enough remaining 7-day allowance.

Choose a larger tier when your measured sessions approach either limit. For example, a plan can have points left in its 5-hour window and still pause because its 7-day allowance is exhausted. Check both balances and reset times in the Dashboard.

When does the standard API make more sense?

Consider the standard API when your usage is intermittent, you need API billing for an application, or a subscription’s usage windows do not suit your workload. Z.ai lists the following model prices:

Official Z.ai API prices were checked October 23, 2026. Tool calls and other services can have separate charges.

For illustration, 1M uncached input tokens plus 200,000 output tokens would cost $2.28 on GLM 5.3 or $0.25 on GLM 5.3 Flash, before any tool charges. This is arithmetic using the listed rates, not a cost estimate for a complete coding project.

To compare a subscription fairly, use the same model, input/output mix and cache assumptions, then check whether your busiest sessions fit its limits. Our Multi-Model plan is worth considering when switching between providers is part of your workflow. It should not be assumed cheaper for every GLM-only workload.

Using GLM in Claude Code or OpenCode

Our existing guides provide separate configurations for each tool. Use the guide for the credential you actually purchased; an official Z.ai key and our gateway key are not interchangeable.

1. Choose the Multi-Model group on Pricing.Check the card’s model lineup, both usage limits, availability and payment mode. The separate DeepSeek Coding Plan has its own key and allowance.
2. Retrieve your delivered credentials.After purchase, use the delivered Multi-Model key and connection details. Our OpenAI-compatible API base is https://multi-api.deepseekv4pro.com/v1.
3. Apply the configuration for your tool.Follow our Claude Code guide or OpenCode guide. Claude Code uses a different protocol configuration; do not paste an OpenAI Base URL into every client.
4. Select a GLM route and verify a small task.Choose glm-5.3-flash or glm-5.3, confirm the selected provider and model, then inspect the result and Dashboard usage before running a longer session.

For official Z.ai access, the Coding Plan endpoint differs from the standard API endpoint. Use the endpoint that matches your account and billing method.

Questions before purchasing