Free Chinese AI API Without a Credit Card

Most AI API providers want a card before you can test anything. NovAI's free tier does not: you can create a key with an email address, call five models for free, and decide later whether paid usage is worth it.

This page is the honest version of what "free" means here — what is $0, what is not, and where the limits are.

What is actually free

Model	Type	Free?
glm-4.7-flash	Text / chat	$0 to call
glm-4.6v-flash	Vision (image input)	$0 to call
glm-4.1v-thinking-flash	Vision + reasoning	$0 to call
cogview-3-flash	Image generation	$0 to call
cogvideox-flash	Video generation	$0 to call

These are Zhipu AI's Flash models served through NovAI. They are genuinely $0 in the price feed, not "free credits that expire". What they do have is the published pre-top-up rate limit: 5 requests per minute before your first top-up. After any completed top-up there is no NovAI-side request cap.

What the $2 signup credit covers

New accounts get $2 of trial credit that can be spent on any paid model, so you can test DeepSeek, Qwen, GLM, Kimi, MiniMax or Doubao output without adding a card. Paid models are billed at the provider list price with a 1.0x multiplier and 0% platform fee — the credit buys $2 of list-price usage, not a marked-up $2.

When the credit and free calls are not enough, top-ups start at $5 (card or USDT TRC20). There are no subscriptions and no per-seat fees.

Get a key in two minutes

1. Open the free tier page and sign up with an email address (no card).
2. Verify the email, then copy the nvai-... key from the dashboard.
3. Point any OpenAI SDK at the endpoint:

from openai import OpenAI

client = OpenAI(
    base_url="https://aiapi-pro.com/v1",
    api_key="nvai-...",
)

resp = client.chat.completions.create(
    model="glm-4.7-flash",   # free model
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(resp.choices[0].message.content)

What you can build with the free tier

• Prototype a chatbot on a free text model, then move to a paid model when quality matters.
• Test vision with glm-4.6v-flash: image understanding without a card.
• Generate one-off images with cogview-3-flash and short video clips with cogvideox-flash.
• Wire an agent framework (LangChain, LlamaIndex, Cline, Continue) and validate your tool loop before paying for tokens.

What "free" does not mean here

To keep the comparison honest:

• It is not unlimited. Free models are rate-limited before your first top-up; the published policy is 5 requests/min.
• It is not every model. The $0 list is the five Flash models above; the other 106 models are paid at list price.
• It is not a separate API. Free and paid models share the same endpoint and key, so moving from free to paid is a model-id change, not a migration.

Check the policy yourself

Everything above is published in machine-readable form:

• /transparency.json — fee, multiplier, free models, minimum top-up, rate-limit note, refund policy.
• /china-ai-leaderboard.json — full price feed, regenerated daily from the model catalogue.
• /free/index.html — the signup flow itself.

If a claim here ever disagrees with the JSON, the JSON wins — that is the point of publishing it.

FAQ

Do I need a credit card to sign up?

No. You can sign up with an email address, get a key, and call the five free models without adding any payment method.

Are the free models really $0?

Yes - glm-4.7-flash, glm-4.6v-flash, glm-4.1v-thinking-flash, cogview-3-flash and cogvideox-flash are $0 in the published price feed. The published pre-top-up rate limit is 5 requests/min.

What can I do with the $2 trial credit?

Spend it on any paid model at provider list price (0% platform fee, 1.0x multiplier) to test DeepSeek, Qwen, GLM, Kimi, MiniMax or Doubao before deciding whether to top up.

What is the minimum top-up?

Paid top-ups start at $5, by card or USDT (TRC20). There are no subscriptions.

Is the free tier a different API?

No. Free and paid models use the same OpenAI-compatible endpoint and the same key; moving to paid is just a model-id change.