Migrate from Azure OpenAI to unoblox (India)

Azure OpenAI works well technically, but for a lot of India-based teams the procurement side is the friction: an Azure subscription, a global billing arrangement, and an invoice that doesn't always land cleanly on an India-only budget. Moving to unoblox keeps the same OpenAI-compatible request shape you already know, while switching billing to rupees on a single monthly GST invoice from an Indian entity.

Why India teams look at moving off Azure OpenAI

The technical capability of Azure OpenAI isn't the issue for most teams — it's everything around it: provisioning an Azure resource before you can call a model, waiting for specific models to reach your region, and reconciling a global billing account against a rupee budget. unoblox skips the provisioning step (sign up, generate a key, start calling) and bills everything on one rupee invoice with GST applied.

What actually changes in your code

Azure OpenAI routes requests through a named "deployment" you create in advance, with an api-version query parameter on every call. unoblox routes by model id directly, with no deployment step and no version parameter.

	Azure OpenAI	unoblox
Routing	Named deployment you provision first	Model id in the request body
Extra setup	Azure resource + deployment per model	None — sign up and generate a key
API version parameter	Required on every request	Not required
Billing	Through your Azure subscription	Monthly ₹ invoice with GST

In practice, this means replacing your deployment name with a model id like openai/gpt-4o or openai/gpt-5, pointing the base URL at unoblox, and dropping the api-version parameter. Everything else about the request — messages array, streaming, tool/function calling, JSON mode — stays the same, because both are OpenAI-compatible at the wire level.

base_url: https://api.unoblox.ai/v1
api_key: ub-gw-xxxxxxxxxxxxxxxxxxxx
model: openai/gpt-4o

What you gain beyond OpenAI's own models

Azure OpenAI only ever gives you OpenAI's models. Behind the same unoblox key, you also get Claude, DeepSeek, Qwen, Llama, Gemma, Kimi, Mistral and Nemotron families — so migrating off Azure isn't just a lateral move, it's a way to add non-OpenAI models to the same codebase without a second SDK or billing relationship.

Migration checklist

1. Sign up at https://unoblox.ai/sign-in and generate a ub-gw-... key.
2. Swap your base URL to https://api.unoblox.ai/v1.
3. Replace each Azure deployment name in your code with the matching model id (for example, your GPT-4o deployment becomes openai/gpt-4o).
4. Remove the api-version query parameter — unoblox doesn't use it.
5. Run your existing test suite against the new endpoint before cutting over production traffic.
6. Keep your Azure resource live in parallel until you've verified output quality and latency on unoblox for your specific prompts.

What doesn't change

Your SDK doesn't change — the standard OpenAI client libraries work against unoblox with just the base URL and key updated. Streaming responses, tool/function calling, and JSON mode all behave the same way, since unoblox implements the same OpenAI-compatible surface your code already targets.

Frequently asked questions

Do I have to rewrite my application to migrate off Azure OpenAI? No — if you're using a standard OpenAI SDK, the change is limited to the base URL, the API key, and swapping deployment names for model ids. The request and response shapes stay the same.

Can I run Azure OpenAI and unoblox side by side during migration? Yes — keep your Azure resource active while you validate output quality, latency, and cost on unoblox, then cut over once confident.

Does unoblox support the same OpenAI models I'm using on Azure? unoblox carries a broad OpenAI model line — GPT-4.1, GPT-4o, GPT-5 and its variants, and the o-series reasoning models — alongside non-OpenAI models. Confirm your specific model id exists on /models before migrating that workload.

Will my invoice look different after migrating? Yes — you'll receive one monthly GST invoice from unoblox in rupees covering every model you call, rather than a bill through your Azure subscription.

Is there a deployment step on unoblox like there is on Azure? No — there's no resource provisioning or deployment step. You generate a key and reference the model id directly in each request.

What if a model I use on Azure OpenAI isn't on unoblox's catalog? Check /models for the current list. If a specific model or version isn't there yet, it's worth flagging to unoblox directly rather than assuming it will never be added.

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