"GLM is free" and "GLM costs $4.40 per million tokens" are both true, and they can both appear on the same search results page. The difference is which GLM you mean and how you plan to use it: downloading the model, calling an API, or using it inside a chat app or coding tool.
Is GLM free? Partly. The model weights are free: GLM-4.6 through GLM-5.2 and GLM-5.3-Flash are MIT-licensed, and GLM-5.3 uses a permissive custom license. Some hosted models are free too: Z.ai's API lists GLM-4.7-Flash, GLM-4.5-Flash and GLM-4.6V-Flash at no charge. The flagship models aren't free to use as a service: GLM-5.x usage on Z.ai's API is billed per token, and its Coding Plan starts at $18 a month.
This guide breaks that down version by version, so you can tell a genuinely free option from a trial that runs out. It's based on Z.ai's official pricing page, Coding Plan documentation, Hugging Face model cards and license files, all reviewed November 19, 2026. We did not test every provider's promotional offer. GLM5 publishes this guide and operates independently of Z.ai.
Is GLM Free? The Answer by Version
| Model | Weights free to download? | Free on Z.ai's API? | Z.ai API price (input / output per 1M tokens) |
|---|---|---|---|
| GLM-5.3 | ✅ Custom GLM-5.3 license | ❌ | $1.40 / $4.40 |
| GLM-5.3-Flash | ✅ MIT | ❌ | $0.15 / $0.50 |
| GLM-5.2 | ✅ MIT | ❌ | $1.40 / $4.40 |
| GLM-5.1 | ✅ MIT | ❌ | $1.40 / $4.40 |
| GLM-5 | ✅ MIT | ❌ | $1.00 / $3.20 |
| GLM-4.7 | ✅ MIT | ❌ | $0.60 / $2.20 |
| GLM-4.7-Flash | ✅ MIT | ✅ Free | Free |
| GLM-4.6 | ✅ MIT | ❌ | $0.60 / $2.20 |
| GLM-4.5-Flash | — | ✅ Free | Free |
| GLM-4.6V-Flash (vision) | — | ✅ Free | Free |
Prices are from Z.ai's pricing page. License information comes from each model's official Hugging Face repository.
Free Option 1: Download the Weights
Every recent flagship GLM model publishes its weights. GLM-4.6, 4.7, 5, 5.1 and 5.2 use the MIT license, which allows free commercial use, modification and redistribution with the license notice kept.
GLM-5.3 is the exception worth knowing about. Its weights ship under a custom GLM-5.3 license. It grants the same broad rights as MIT, with one added condition: a company running a "model as a service" business with more than $10 billion in revenue over 12 months must pass Z.ai's security review before commercial use. For nearly everyone that condition never applies, but it means GLM-5.3 isn't technically MIT.
The catch with free weights is hardware. Flagship GLM checkpoints are hundreds of billions of parameters. GLM-5.2's is roughly 753 billion. Running them takes multi-GPU servers, which are expensive to buy or rent. "Free to download" is not "free to run."
The practical exception is GLM-4.7-Flash. It's a 30B-total, 3B-active model, small enough for far more modest hardware and also MIT-licensed. If you want a GLM model on your own machine at no software cost, it's the realistic starting point.
Free Option 2: Z.ai's Free API Models
Z.ai's pricing page lists three models as completely free on its API: input, cached input and output all at zero.
- GLM-4.7-Flash: the lightweight GLM-4.7 variant, with a 200K-token context per Z.ai's documentation.
- GLM-4.5-Flash: an older lightweight model.
- GLM-4.6V-Flash: a free vision model that accepts images.
These are good for prototypes, personal projects, learning the API and light production tasks. They aren't the flagships. Expect them to trail GLM-5.x on hard coding and long agent tasks.
What Isn't Free: Flagship API Usage
GLM-5, 5.1, 5.2 and 5.3 are paid per token on Z.ai's API. A rough sense of scale at GLM-5.2/5.3 rates ($1.40 input, $4.40 output per million tokens):
| Usage | Approximate cost |
|---|---|
| A short chat exchange (2K tokens in, 1K out) | Under one cent |
| A 100K-token document plus a 2K-token answer | About $0.15 |
| A full 1M-token prompt plus a 10K-token answer | About $1.44 |
Cached input, which is billed at $0.26 per million on Z.ai, makes repeated long prompts much cheaper.
What About the GLM Coding Plan?
Z.ai's GLM Coding Plan is a subscription for using GLM inside coding tools such as Claude Code, Cline and OpenCode. Z.ai's documentation says plans start at $18 per month, with higher Pro and Max tiers. It isn't free. Z.ai has offered trial cards, but treat those as time-limited.
One detail affects what you get: on the Coding Plan, requests for GLM-5.2 or 5.1 are routed to GLM-5.3, and GLM-4.7 requests are routed to GLM-5.3-Flash, per the Coding Plan overview.
Free GLM in a Browser
If you just want to try GLM without hardware or an API key, browser chat is the simplest path. On GLM5, the free Chat default is GLM 5.3 Flash, and other GLM models can be selected with credits. Open GLM5 Chat to try it, and check the model selector to see which GLM version is answering.
Free Answers for Specific Versions
- Is GLM 5 free? The weights are free under MIT. Z.ai's API charges $1.00 input and $3.20 output per million tokens. See what GLM-5 is.
- Is GLM 5.1 free? The weights are free under MIT. API use is $1.40/$4.40. See what GLM 5.1 is.
- Is GLM 5.2 free? The weights are free under MIT, and API use is paid. Our detailed GLM 5.2 guide covers trials and providers.
- Is GLM 5.3 free? The weights are free under the GLM-5.3 license. API use is $1.40/$4.40, and GLM-5.3-Flash is the cheaper tier at $0.15/$0.50.
- Is GLM 4.7 free? The weights are free. GLM-4.7 itself is paid on the API, but GLM-4.7-Flash is free. See what GLM 4.7 is.
How to Pick the Right "Free" Option
| Your goal | Best free (or near-free) path |
|---|---|
| Try GLM today, no setup | Browser chat |
| Build a prototype with an API | GLM-4.7-Flash on Z.ai's free tier |
| Run GLM on your own computer | GLM-4.7-Flash weights |
| Image input for free | GLM-4.6V-Flash on Z.ai's free tier |
| Flagship quality for real work | Paid API usage, starting small with token caps |
Rule of thumb: use free tiers to find out whether GLM fits your task. Pay for a flagship only once a free model has clearly failed on your real prompts.
Frequently Asked Questions
Is GLM free to use?
Partly. The weights are free to download, and Z.ai's API offers GLM-4.7-Flash, GLM-4.5-Flash and GLM-4.6V-Flash for free. Flagship GLM-5.x models are paid per token.
Is GLM open source?
The weights are open: MIT for GLM-4.6 through GLM-5.2, and a permissive custom license for GLM-5.3. See what GLM is for background.
What is the best free GLM model?
GLM-4.7-Flash. It's free on Z.ai's API, MIT-licensed, and small enough (30B total, 3B active) to self-host on modest hardware.
Is the GLM Coding Plan free?
No. Plans start at $18 per month, according to Z.ai.
Can I use free GLM models commercially?
MIT-licensed weights allow commercial use with the license notice kept. For hosted free tiers, check Z.ai's terms for rate limits and usage conditions.
The Bottom Line
GLM is free in two real ways: open weights you can download, and a handful of Flash models Z.ai serves at no charge. It isn't free in the way most people hope, because the flagship GLM-5.x models cost money to run, whether you pay per token, subscribe to a coding plan or buy the GPUs yourself.
The smart path is to start free and pay only when you need to: try GLM in your browser on GLM5, test GLM-4.7-Flash on Z.ai's free tier, and move to a flagship once your real tasks prove they need it.
Sources
- Z.ai API pricing — free Flash models and per-token prices for every GLM version.
- GLM-5.3 license — custom license terms and the security-review condition.
- GLM-4.7-Flash official model card — 30B-A3B size and MIT license.
- GLM Coding Plan overview — plan pricing and model routing.
Reviewed November 19, 2026. Free tiers, prices and plan terms change often. Confirm them on Z.ai's official pages before you rely on them.



