OpenAI o3 API in India

o3 is OpenAI's reasoning-focused model, built to work through multi-step problems — math, code, planning — before answering, and it's available in India through unoblox on the same ₹-billed, OpenAI-compatible endpoint as the rest of the catalogue, with one monthly GST invoice replacing a foreign-currency bill.

What makes a reasoning model different to call

Reasoning models like o3 spend part of their output budget on internal reasoning before producing the final answer. In practice this means two things worth planning for: responses can take longer to return than a same-sized non-reasoning model, and the max_tokens (or equivalent completion-token) budget needs enough headroom for that internal reasoning plus the visible answer, or the response can get cut off before it reaches a conclusion. Neither of these is specific to unoblox — it's how o-series reasoning models behave everywhere.

₹ pricing for o3

o3 doesn't have a fixed indicative rate in this note — its live ₹ price is on the model page, and it's worth checking before committing a workload to it, since reasoning models can consume more output tokens per answer than a standard chat model for the same question. For rough context, here's where OpenAI's flagship non-reasoning tier sits:

Model	Input (₹ / 1M tokens)	Output (₹ / 1M tokens)
GPT-5	126	1008
GPT-4.1	201.6	806.4

Calling o3 from the unoblox endpoint

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/o3",
    "messages": [{"role": "user", "content": "Work through this multi-step pricing calculation and show your steps."}],
    "max_tokens": 1200
  }'

Giving the request a generous completion-token budget, as above, is the single most common fix when a reasoning-model response comes back empty or truncated.

When o3 is worth reaching for

• Problems with several dependent steps — multi-part calculations, debugging a stack trace against source, planning a sequence of tool calls.
• Tasks where a wrong intermediate step ruins the final answer, so the extra reasoning time is worth paying for.
• Not a fit for high-volume, low-complexity calls — route those to a mini or nano-class model instead and reserve o3 for the subset that actually needs it.

Frequently asked questions

Is o3 available in India through unoblox? Yes, openai/o3 is callable on https://api.unoblox.ai/v1 immediately after signup, billed in rupees against wallet balance.

What does o3 cost on unoblox? It isn't in our fixed indicative price list; see the live ₹ rate on its model page before estimating a workload's cost.

Why did my o3 response come back empty or cut off? Reasoning models use part of the completion-token budget internally before writing the visible answer — raise max_tokens to give it room to finish.

Does o3 run on infrastructure located in India? No, it runs on OpenAI's infrastructure. The India-specific benefit from unoblox is ₹ billing, a GST invoice and one endpoint, not data residency.

Can I call o3 and a non-reasoning model like GPT-5 with the same key? Yes, the same ub-gw- key works across every model in the catalogue.

Is GST input tax credit available on o3 usage? Yes, o3 usage is included in the same monthly GST invoice as every other model, and is claimable as input tax credit.

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