Free LLM API — no card required

Free LLM API, no payment wall

Most "free tiers" quietly demand a card on file before you can call the API — the free part only kicks in after that gate. unoblox's free tier (Qwen3.8-27B) skips the gate entirely: no card, no international payment method, no expiring trial. You sign up with email and start calling the endpoint.

What "free" hides elsewhere, and what it means here

• Card-gated free tiers: you still need a payment method capable of international billing before the free credits activate.
• Time-boxed trials: free for 14 or 30 days, then a card is mandatory.
• Free-but-metered surprises: a generous quota that quietly starts billing once you cross it, sometimes without a clear warning.

unoblox's free tier avoids all three: ₹0, no card, no countdown — just a fair-use rate limit so the tier isn't abused for resale.

The model behind it

Model: Qwen3.8-27B, Alibaba's open-weight 27B model with text and image input. Endpoint: the same OpenAI-compatible /v1/chat/completions used by every other model on unoblox — no separate SDK. Pricing: ₹0 input, ₹0 output. free calls cost ₹0, so they never reduce your wallet balance. Rate limit: fair-use limit of 10 requests per minute per workspace. There is no monthly cap.

Quickstart (no card, three steps)

1. Sign up at https://unoblox.ai/sign-in with email and password.
2. Create an API key (ub-gw-…) from the dashboard.
3. Call it:

from openai import OpenAI

client = OpenAI(api_key="ub-gw-YOUR_KEY", base_url="https://api.unoblox.ai/v1")
resp = client.chat.completions.create(
    model="qwen/qwen3.8-27b",
    messages=[{"role": "user", "content": "Summarize this in two lines: ..."}]
)
print(resp.choices[0].message.content)

Free calls cost ₹0 — nothing to reconcile, nothing to cancel.

Free tier vs a paid model, what actually changes

Dimension	Free (Qwen3.8-27B)	Paid (e.g. Llama 4 Scout)
Cost per 1M tokens	₹0 / ₹0	₹10.12 / ₹30.36
Context window	262,144 tokens	128K tokens
Best for	Classification, short generation, prototyping	Higher-accuracy production workloads
Support	Community, docs	Email support
Rate limit	10 req/min per workspace (fair use)	Set by your account

Nothing else changes — same endpoint, same key format, same SDK. Switching later is a one-line model change.

When free genuinely stays free

Prototyping, internal tools, low-traffic classification or summarization endpoints, and hackathon-style projects rarely approach the rate limit. If your usage is bursty rather than constant, the free tier can outlast the project itself.

When to move off it

Move to a paid model once you need: consistently higher accuracy on open-ended tasks, a longer context window than 262K tokens, or predictable throughput above the shared rate limit. All three are one config change away — no re-integration.

Frequently asked questions

Why is this free if the model is genuinely useful? Small instruction-tuned models are cheap enough to run at scale that unoblox can offer them at ₹0 without subsidizing a loss — it's also how new users try the platform before committing spend to bigger models.

Is there a sunset date for the free tier? No planned one. unoblox reserves the right to introduce stricter abuse limits if the tier is misused, but the ₹0 price itself isn't time-limited.

Can I use it for a commercial product? Yes, within the fair-use rate limit. Many early-stage products run real traffic on it.

Do I need a business account or GSTIN to use the free tier? No. A personal email is enough to start; add GST details later if you upgrade and need ITC-eligible invoices.

What counts as abuse? Reselling API access, running synthetic content farms, or scripted scraping at volume. Normal application traffic, even fairly busy traffic, isn't a problem.

Can I migrate from another provider's free tier to this one? Yes — it's a two-line change: swap base_url to https://api.unoblox.ai/v1 and use your ub-gw-… key.

Get started in rupees → https://unoblox.ai/sign-in