OpenAI-compatible API in India (₹ pricing)

One API endpoint for every model, ₹-native billing

unoblox is a fully OpenAI-compatible endpoint at /v1. Swap your base URL and API key — no other code changes — and run GPT-5, Claude, Qwen, or DeepSeek on one gateway, billed monthly in rupees with GST input tax credit claimable.

Python example: 2-line change

from openai import OpenAI

client = OpenAI(
    base_url="https://api.unoblox.ai/v1",  # change #1
    api_key="ub-gw-YOUR_KEY"                # change #2
)

response = client.chat.completions.create(
    model="qwen/qwen3-235b-a22b-instruct-2507",
    messages=[{"role": "user", "content": "Explain GST in India"}]
)
print(response.choices[0].message.content)

Same chat.completions, same stream=True, same error classes — nothing else in your code needs to change.

Full API parity

Feature	unoblox
/v1/chat/completions (streaming)	Yes
/v1/embeddings	Yes
/v1/rerank	Yes (native, not part of every OpenAI-compatible spec)
JSON mode	Yes
Tool / function calling	Yes
Vision	Yes (Qwen, Claude, GPT)

Which model to pick

• Fast tasks (summarize, classify): Qwen3.8-27B — ₹16.32 / ₹48.96 per 1M
• Deep reasoning: DeepSeek V3.2 — ₹26.21 / ₹38.30 per 1M
• Production quality: Claude Sonnet — ₹201.6 / ₹1008 per 1M
• Free tier: Qwen3 1.7B — ₹0, rate-limited monthly caps

Works from any client, not just Python

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o",
    "messages": [{"role": "user", "content": "Explain GST in India"}]
  }'

Because unoblox speaks plain HTTP and JSON, any language with an HTTP client works — curl, Node's fetch, Go's net/http, Java's HttpClient — even without an OpenAI SDK. The Python and Node.js official SDKs are simply the most convenient wrapper.

Migrating an existing OpenAI integration — checklist

1. Replace base_url / baseURL with https://api.unoblox.ai/v1
2. Replace the API key with your ub-gw-… key
3. Replace the model string with a unoblox model id from /models
4. Re-run your test suite — request and response shapes are unchanged, so assertions on message content still pass
5. Point finance at your unoblox dashboard for the new monthly ₹ invoice

Frequently asked questions

Will my code break when I swap endpoints? No. The SDK is just an HTTP client — swap base_url and api_key; everything else behaves the same.

Do I need to rewrite async code? No. Async clients and await calls work identically, including streaming.

Can I use multiple models in one app? Yes. Instantiate one client and pass different model ids per call — each is priced independently.

How do I handle rate limits? Same as any OpenAI-compatible API: catch a 429, back off, and retry.

What if a model isn't available on unoblox? Check /models for the live catalog — new models are added regularly, and it's the source of truth over any hardcoded list.

Am I locked into unoblox if I use the OpenAI SDK? No. The SDK just speaks HTTP to whatever base_url you configure — you can point it elsewhere at any time.

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