Yes — there’s a genuinely free way to call a GLM model through an API, but “free” splits into three different deals depending on which channel you use and which GLM version you mean. Z.ai’s own API gives permanent free access to two smaller GLM models with no rate-limit expiration. OpenRouter gives free, rate-limited access to GLM-5.2 specifically. And GLM-4.5-Air, the version most people search for by name, isn’t free anywhere — it’s a low-cost paid model, and any page telling you otherwise is out of date.
| Channel | Free form | Models covered |
|---|---|---|
| z.ai official API | Permanent free tier, no expiration | GLM-4.7-Flash, GLM-4.5-Flash |
OpenRouter :free tag | Rate-limited free tier (20 req/min, 50-1,000 req/day) | GLM-5.2 only |
| chat.z.ai (browser chat, not API) | Free web chat, no API key | GLM-5.2 (chat UI, no programmatic access) |
The Three Free Channels, in Detail
z.ai’s Own API: Two Models Are Actually Free
Z.ai’s official pricing page lists GLM-4.7-Flash and GLM-4.6V-Flash and GLM-4.5-Flash at $0 for input, cached input, and output — not a discount, not a trial, just free. Every other model in z.ai’s lineup, including GLM-4.5, GLM-4.5-Air, GLM-5.2, and GLM-5.3, carries standard per-token pricing. (Checked 2026-08-25 via docs.z.ai/guides/overview/pricing.)
This is the cleanest “free” deal on this list because it’s not rate-limited by request count and doesn’t require a credit card. The catch is model choice: Flash-tier models trade off some capability for cost, so if your workload specifically needs GLM-4.5-Air’s reasoning quality or GLM-5.2’s long-context handling, the free Flash models aren’t a substitute — they’re a different tier of model.
OpenRouter: Free, but Only for GLM-5.2, and Rate-Limited
OpenRouter runs GLM-5.2 under a :free suffix (z-ai/glm-5.2:free) at $0 per token for both input and output. Per OpenRouter’s own rate-limit documentation, any model ending in :free is capped at 20 requests per minute; the daily cap depends on how much you’ve spent on the platform lifetime — under $10 in lifetime credits gets you 50 requests per day, and a one-time $10 top-up (which never expires) raises that to 1,000 requests per day. (Checked 2026-08-25 via openrouter.ai/docs/api-reference/limits and confirmed against the live model list at openrouter.ai/api/v1/models.)
GLM-4.5-Air does not have a free tier on OpenRouter as of this check — it’s listed there as a standard paid model. If you’ve seen a guide referencing glm-4.5-air:free, that variant existed at one point but isn’t in OpenRouter’s current model catalog; treat any tutorial still pointing at it as stale. GLM-5.2 is the GLM model currently carrying the :free tag on this platform.
chat.z.ai: Free, but It’s a Chat UI, Not an API
Z.ai also runs a free browser chatbot at chat.z.ai, currently serving GLM-5.2, with no sign-in or API key required to start chatting. (Checked 2026-08-25: the page’s title tag and meta description both read “powered by GLM-5.2”.) This is worth knowing about if what you actually want is to try a GLM model’s output quality before writing any code — but it’s not programmatic access. There’s no API key, no request format, no way to call it from your own application. If your goal is an API integration, this channel doesn’t apply; use one of the two above instead.
Why GLM-4.5-Air Isn’t on This List as “Free”
If you searched for a free GLM API and landed here specifically wondering about GLM-4.5-Air, the direct answer is: it’s not free, on z.ai, OpenRouter, or Novita AI. Current verified pricing:
| Provider | GLM-4.5-Air input (per 1M tokens) | GLM-4.5-Air output (per 1M tokens) |
|---|---|---|
| z.ai (official) | $0.20 | $1.10 |
| Novita AI | $0.13 | $0.85 |
(Checked 2026-08-25 via docs.z.ai/guides/overview/pricing and api.novita.ai/openai/v1/models.)
GLM-4.5-Air is a real model with a real cost on every platform that serves it — the “Air” in the name refers to a smaller, cheaper variant of GLM-4.5, not a free one. Its actual position is: cheap relative to flagship models like GLM-5.2 or GLM-5.3, but not zero-cost. If you need GLM-4.5-Air specifically, the free channels above won’t get you there; you’re picking a provider based on price and reliability, not searching for a free tier that doesn’t exist.
Novita AI hosts the same GLM-4.5-Air model at a lower per-token rate than z.ai’s own API, with an OpenAI-compatible endpoint (no separate SDK or model-specific integration work if you’re already calling other models through the same client). That’s a fit for teams that want GLM-4.5-Air in production without a rate-limited free tier standing in the way, not a claim that it’s free.
Separately, Novita AI has run its own zero-cost GLM offering before — but that was GLM-4-9B-0414 and GLM-Z1-9B-0414, two smaller 9B-parameter models from an earlier GLM generation, not the GLM-4.5+ series people are searching for today. If you find that older free-model page, know that it’s a different pair of models with different specs, not a free tier for GLM-4.5-Air.
Which One Should You Actually Use?
If you specifically need GLM-4.7-Flash or GLM-4.5-Flash and those models fit your task, z.ai’s own API is the simplest free option — no rate-limit ceiling, no credit card. If you need GLM-5.2 and can live with 50 requests a day (or want to spend $10 once to get 1,000/day), OpenRouter’s free tier works and gives you a single API that also covers other providers if you want to compare models later. If you need GLM-4.5-Air, GLM-4.5, GLM-4.6, or any of the flagship GLM-5.x models, there’s no free channel for those — compare per-token pricing across providers instead, since that’s the actual lever you control.
FAQ
Is GLM-4.5-Air free on Novita AI?
No. It’s priced at $0.13 per 1M input tokens and $0.85 per 1M output tokens on Novita AI as of this check, and $0.20/$1.10 on z.ai’s own API. No provider currently serves GLM-4.5-Air for free; if you’ve read otherwise, check the publish date on that source, since GLM pricing has shifted across versions.
Is GLM-5.2 free?
It’s free on OpenRouter under the :free tag, subject to the 20 requests/minute and 50-or-1,000 requests/day caps described above. On z.ai’s own API and on Novita AI, GLM-5.2 is priced normally — the free access is specific to OpenRouter’s free-tier program, not a universal free release of the model.
What’s the actual difference between z.ai’s free Flash models and OpenRouter’s free GLM-5.2?
Z.ai’s free models (GLM-4.7-Flash, GLM-4.5-Flash) have no request-count ceiling and don’t expire, but they’re the smaller, faster tier rather than the flagship model. OpenRouter’s free GLM-5.2 gives you the larger, more capable model, but caps you at a daily request count. Pick based on whether your constraint is model capability or request volume.
Can I get GLM-4.5-Air access without paying per token?
Not currently, on any provider checked here. Your options are z.ai’s official pricing, Novita AI’s lower per-token rate for the same model, or downloading GLM-4.5-Air’s open weights and self-hosting on your own GPU, which removes the per-token bill but replaces it with infrastructure cost.
