Migrate from OpenAI to unoblox

Why teams migrate from OpenAI to unoblox

OpenAI's API is reliable, but it bills internationally and is not an Indian entity — friction that shows up at the finance and compliance stage even when the product experience is fine. unoblox mirrors OpenAI's entire surface — GPT-4o, GPT-4.1, GPT-5, streaming, tools, vision, embeddings — behind one ₹-invoiced endpoint from an Indian entity.

The migration is one line

Your OpenAI SDK code is reusable as-is. Only the endpoint and key change:

# BEFORE (OpenAI)
from openai import OpenAI

client = OpenAI(api_key="sk-...")
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}]
)

# AFTER (unoblox) — same code, different endpoint + key
client = OpenAI(
    api_key="ub-gw-...",   # from admin.unoblox.ai
    base_url="https://api.unoblox.ai/v1"
)
# Streaming, tools, JSON mode, embeddings — all unchanged

That's the entire migration: no rewrite, no library swap.

Real ₹ pricing

Model	₹ input / 1M	₹ output / 1M
GPT-5	₹126.53	₹1,012.28
GPT-5 mini	₹25.30	₹202.46
GPT-4o	₹253.07	₹1,012.28
GPT-4.1	₹202.46	₹809.82
Embeddings (Qwen3)	₹0	—

Every model above is paid from the same rupee wallet, with a GST tax invoice issued at each top-up — and every rupee of it is eligible for 18% input-tax credit.

Features that carry over unchanged

• Chat completion streaming
• Tool calls (function calling)
• JSON mode
• Vision (image understanding) on GPT-4o and Qwen3.8-27B
• Embeddings and vector search
• The exact request/response schema your app already parses

What you gain by switching

• No international payment step. Prices are quoted and billed in rupees, full stop.
• A GST invoice your accountant recognizes. Input-tax credit claims stop being a special case.
• Indian entity. Your account data, logs and GST invoices are with an Indian company. GPT models still run on OpenAI's infrastructure outside India; only unoblox's self-hosted models run in India.
• More models on the same key. Claude, DeepSeek, and Qwen sit next to GPT without a second contract.
• Same-day invoicing, with payment terms negotiable at volume.

A simple migration checklist

1. Create an unoblox account and copy your ub-gw-… key.
2. Swap base_url in your OpenAI client — no other code changes.
3. Run your existing test suite against the new endpoint.
4. Shadow-traffic a percentage of production requests for a few days.
5. Cut over fully once error rates and latency look right.

Frequently asked questions

Will my application break? No — unoblox matches the OpenAI API surface, so anything that works against OpenAI works here once the endpoint and key change.

Can I keep OpenAI as a fallback? Yes. Store both api_key and base_url pairs in config and add retry-to-fallback logic while you build confidence in the new endpoint.

What happens to my OpenAI conversation history? It stays on OpenAI's servers — unoblox starts fresh, since neither platform exports conversation state today.

Does tool calling work exactly the same? Yes — pass the same tools array and read tool_calls from the response exactly as you do today.

How are rate limits handled? Your account tier sets RPM/TPM. For high-volume launches, talk to sales@unoblox.ai about reserved capacity ahead of time.

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