api.z.ai
Free LLM quota · free models · community reputation · risk review
Overview
The API domain for Z.ai, Zhipu AI's international developer platform (api.z.ai 301-redirects to z.ai/model-api). It sells pay-as-you-go access to the in-house GLM model family, with the GLM-Flash text and vision models offered free on an ongoing basis.
Free quota
GLM-4.7-Flash, GLM-4.5-Flash (text) and GLM-4.6V-Flash (vision) are listed at zero cost for both input and output on the official pricing page, making them ongoing free models rather than trial credits. The free tier is rate-limited (third-party measurements suggest roughly 1 QPS and on the order of 1000 calls/day; no official figure is published). Flagship GLM-5.x models are paid, and a separate GLM Coding Plan subscription exists.
Free models
GLM-4.7-FlashGLM-4.5-FlashGLM-4.6V-FlashPros
- A genuinely callable API: OpenAI-compatible endpoint (base_url of the form https://api.z.ai/api/paas/v4) usable directly from the openai SDK
- The Flash line is priced at zero on an ongoing basis, not a time-limited trial allowance
- Run directly by the model vendor (Zhipu AI) rather than a relay, so there is no upstream-key-theft concern
- Simple signup: an email is enough to generate an API key in the console, no credit card required
- Covers text, vision and video generation, is listed on AWS Marketplace, and is natively supported by gateways such as liteLLM
Cons
- Only the small Flash models are free; every flagship GLM-5.x model is paid
- Free-tier rate and concurrency ceilings are not published; third-party measurements put it near 1 QPS and ~1000 calls/day, unusable under real load
- Flash models are tuned for speed and reason less deeply than the flagship tier
- The paid GLM Coding Plan draws recurring rate-limit complaints; the vendor adjusts concurrency dynamically and publishes no concrete SLA numbers
- The host recorded here is the API domain api.z.ai, which 301-redirects to z.ai/model-api; the real user entry point is z.ai
- The China-domestic platform (bigmodel.cn) and the international one (z.ai) have separate accounts, quotas and legal entities, which is easy to confuse
Risk flags
- Concrete RPM and daily-call ceilings for the free tier are not published, so limits rest on third-party measurement and can tighten at any time
- Only the Flash line is free; the vendor can reprice it out of the free tier at any point
- The paid Coding Plan has documented complaints about dynamic throttling with no published concurrency SLA
- Operated by Zhipu AI, a mainland China vendor, so data handling follows Chinese law; the international and domestic entities have different terms and must be checked separately
- Documentation does not state whether free-tier traffic is used for training; do not send sensitive data
- api.z.ai is an API-only domain that redirects, so verify the domain carefully against phishing lookalikes
Community reputation
Z.ai is Zhipu AI's international brand, a top-tier Chinese model vendor with a clear corporate entity. Public discussion centres on model capability benchmarks (GLM-4.5/4.6/4.7 on coding and agentic tasks) and cheap subscriptions, and is broadly favourable; no exit-scam or fraud allegations were found. The main negative thread concerns the paid GLM Coding Plan's rate limiting: the vendor never published a concurrency figure and tightens limits dynamically based on resource availability, which users experience as paying for throughput they cannot reach. The free Flash tier itself attracts little controversy but also lacks published quota documentation.
Recommendation
This is a vendor free tier. The Flash models are genuinely priced at zero and programmatically callable, which makes them reasonable for low-cost chat, classification and summarisation prototypes plus light vision work. Do not build high-concurrency or deep-reasoning production paths on it: the free ceiling is undocumented and covers only small models, and flagship GLM-5.x still costs money. Sign up via z.ai rather than api.z.ai, and note that accounts on the domestic bigmodel.cn platform do not carry over. Keep sensitive data off the free tier.
Latest probe
Mentions (5)
Sources
Community discussion · reputation check
Want to verify api.z.ai's real reputation before deciding? Read first-hand community discussion:
External discussion quality varies; judge together with the risk review on this page.
FAQ
Is api.z.ai safe and trustworthy?
Overall reputation 76/100, risk level Low. Z.ai is Zhipu AI's international brand, a top-tier Chinese model vendor with a clear corporate entity. Public discussion centres on model capability benchmarks (GLM-4.5/4.6/4.7 on coding and agentic tasks) and cheap subscriptions, and is broadly favourable; no exit-scam or fraud allegations were found. The main negative thread concerns the paid GLM Coding Plan's rate limiting: the vendor never published a concurrency figure and tightens limits dynamically based on resource availability, which users experience as paying for throughput they cannot reac…
Will api.z.ai shut down? Is the service stable?
Known risks: Concrete RPM and daily-call ceilings for the free tier are not published, so limits rest on third-party measurement and can tighten at any time; Only the Flash line is free; the vendor can reprice it out of the free tier at any point; The paid Coding Plan has documented complaints about dynamic throttling with no published concurrency SLA; Operated by Zhipu AI, a mainland China vendor, so data handling follows Chinese law; the international and domestic entities have different terms and must be checked separately; Documentation does not state whether free-tier traffic is used for…
Which free models does api.z.ai offer? How much free quota?
Free quota: GLM-4.7-Flash, GLM-4.5-Flash (text) and GLM-4.6V-Flash (vision) are listed at zero cost for both input and output on the official pricing page, making them ongoing free models rather than trial credits. The free tier is rate-limited (third-party measurements suggest roughly 1 QPS and on the order of 1000 calls/day; no official figure is published). Flagship GLM-5.x models are paid, and a separate GLM Coding Plan subscription exists.. Available free models: GLM-4.7-Flash, GLM-4.5-Flash, GLM-4.6V-Flash.
Is api.z.ai worth using? How should I use it?
This is a vendor free tier. The Flash models are genuinely priced at zero and programmatically callable, which makes them reasonable for low-cost chat, classification and summarisation prototypes plus light vision work. Do not build high-concurrency or deep-reasoning production paths on it: the free ceiling is undocumented and covers only small models, and flagship GLM-5.x still costs money. Sign up via z.ai rather than api.z.ai, and note that accounts on the domestic bigmodel.cn platform do not carry over. Keep sensitive data off the free tier.
Related free-LLM sites
- huggingface.co — Official (free tier) · reputation 90
- console.groq.com — Official (free tier) · reputation 90
- console.mistral.ai — Official (free tier) · reputation 88
- developers.cloudflare.com — Official (free tier) · reputation 88
- aws.amazon.com — Official (free tier) · reputation 85
- ai.google.dev — Official (free tier) · reputation 84
- bailian.console.alibabacloud.com — Official (free tier) · reputation 84
- console.scaleway.com — Official (free tier) · reputation 84