GLM Coding Plan: Pricing, Credit Limits & Claude Code Setup

Update (9 Nov 2026): Claude Fable 5.2 and Mythos 5.2 are out - prices, access and benchmarks are in my Claude Fable 5.2 guide.

👁 119 reads

GLM Coding Plan is Z.ai’s subscription for coding with GLM-5.3 and GLM-5.3-Flash models in Claude Code, Cline, OpenCode, and several other tools. It starts at $18 per month and since July 30, 2026, uses a credit system: Lite gives you 2,000 credits per 5 hours and 10,000 per week. Below we calculate what that means in tokens, how to connect the plan to Claude Code, and what to watch out for. Everything about Z.ai models is in our GLM guide.

Plans and Limits

Credits, token estimates, and recommendations according to Z.ai documentation, October 3, 2026. Lower token bound assumes peak hours, upper bound off-peak. Documentation only lists the starting price of $18 - Pro and Max prices are visible only after logging into the subscription page. External sources report $80 for Pro and $168 for Max monthly, plus discounts for annual payment, but we haven’t verified these figures at the source. You pay by card or PayPal, in USD.

The 5-hour limit renews 5 hours after depletion, the weekly limit every 7 days from purchase date. Account sharing is prohibited by the terms of service.

How Credits Are Calculated

Formula from documentation: credits = (input tokens × input multiplier + cached tokens × cache multiplier + output tokens × output multiplier) / 10,000.

Outside peak hours, everything costs half the credits. Peak is Monday-Friday 14:00-18:00 Singapore time, which in October is 6:00-10:00 UTC. For someone coding in the afternoon and evening in US or European time zones, almost all work falls outside peak hours.

Example from our test of Chinese models in Polish: eight queries to GLM-5.3 consumed 1,904 input tokens and 4,556 output tokens. In plan credits, that’s (1,904 × 6.9 + 4,556 × 24) / 10,000 = 12.2 credits at peak and 6.1 off-peak. The 5-hour limit in Lite plan is 2,000 credits. Output is almost 3.5 times more expensive than input, so an agent that reads a lot but writes little lasts longer.

How to Connect GLM Coding Plan to Claude Code

1. Purchase a plan on the Z.ai subscription page and create an API key in the console.
2. Install Claude Code if you don’t have it (instructions in our Claude Code guide).
3. Open the file ~/.claude/settings.json (on Windows: C:UsersYour_name.claudesettings.json) and add to the env section:
   • "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic"
   • "ANTHROPIC_AUTH_TOKEN": "your_z.ai_key"
   • "API_TIMEOUT_MS": "3000000" - longer timeout because GLM takes time on long tasks.
4. Optionally map models to roles: Z.ai documentation suggests glm-5.3[1m] as the model replacing Sonnet and Opus, and glm-5.3-flash[1m] replacing Haiku. The [1m] suffix indicates 1 million token context.
5. Launch a new terminal and type claude. From now on, queries go to Z.ai and consume plan credits.

Z.ai also has a tool that sets this up automatically, but we prefer to know what changes in our files. For other tools (Cline, OpenCode, Kilo Code), the URL is different: https://api.z.ai/api/coding/paas/v4. The plan key only works in supported tools - you can’t hook it up to your own script like a regular API.

GLM Coding Plan vs. Pay-per-Token API?

In Z.ai’s API, GLM-5.3 costs $1.40 per million input tokens, $0.26 for cache reads, and $4.40 per million output tokens. A coding agent reads heavily, and most input hits the cache. According to Z.ai’s estimate, Lite plan gives you 48-97M GLM-5.3 tokens per week with 95% cache hits. For the same pool in the API, you’d pay according to the price list from around $15 to over $30 weekly, and that’s without output tokens - so for daily work, the subscription is cheaper.

The API makes sense when you call GLM from your own program, need other models (e.g., GLM-4.6V for images), or code a few hours per month. Prices for all models are in the table in our GLM guide.

What to Watch Out For

• No refunds. According to Z.ai’s FAQ, subscriptions are non-refundable, even if unused. Start with a month, not a year.
• Policy changes. On July 30, 2026, Z.ai switched from query limits to credits, and old plans disappeared from sale. Existing subscribers kept their terms until the end of their billing period.
• Data. The team plan states that code and commands don’t go to model training by default. For individual plans, we didn’t find such a declaration in the plan documentation. About what Z.ai’s own application did with code, we wrote in 564 attempts to send 313 MB of code to a Chinese server, and general principles are in coding agent in the enterprise.
• Polish in comments and documentation. We tested GLM-5.3 in Polish and it writes with errors. In code that doesn’t bother us, but in a README for a client it does.

Alternatives

More tools and price comparison: AI for programming and Kimi Code.

Frequently Asked Questions

How much does GLM Coding Plan cost?

From $18 per month for Lite plan. Z.ai shows Pro and Max prices after login; external sources report $80 and $168 monthly.

How many tokens does GLM Coding Plan Lite give you?

2,000 credits per 5 hours and 10,000 per week. According to Z.ai, that’s 48-97M GLM-5.3 tokens weekly with 95% cache hits.

What models are in GLM Coding Plan?

GLM-5.3 and GLM-5.3-Flash. Older model names are automatically redirected to these two.

Does GLM Coding Plan work with Claude Code?

Yes. You set ANTHROPIC_BASE_URL to https://api.z.ai/api/anthropic and your Z.ai console key as ANTHROPIC_AUTH_TOKEN.

Can you refund GLM Coding Plan?

No. According to Z.ai’s FAQ, subscriptions are non-refundable. You must cancel at least 24 hours before the next payment; the plan works until the end of the paid period.

When do credits drain slower?

Outside peak hours (peak: Monday-Friday 14:00-18:00 Singapore time), consumption is half. Until October 7, 2026, this rate applies around the clock.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and Verification Date

Z.ai documentation

• GLM Coding Plan overview
• team plan
• terms of use
• FAQ
• Claude Code configuration
• July 30
• 2026 plan change announcement
• and GLM-5.3-Flash promotion terms. API pricing: docs.z.ai pricing. Token consumption: our test from October 3
• 2026. Verified October 3
• 2026