OpenAI o1 API in India

o1 is OpenAI's original o-series reasoning model, built to work through a problem internally — step by step — before writing a final answer, 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.

How a reasoning model like o1 behaves differently

o1 spends part of its output budget on internal reasoning steps before it writes the answer you actually see. Two practical consequences follow: responses can take longer to come back than from a same-sized non-reasoning model, and the completion-token budget needs enough room for that internal reasoning plus the visible answer — too small a max_tokens and the response can be cut off before it reaches a conclusion. As a rule of thumb, give o1 requests at least 1000 completion tokens of headroom, more for genuinely multi-step problems.

₹ pricing for o1

o1 doesn't have a fixed indicative rate in this note — see the live price on its model page before committing a workload to it, since reasoning models can use more output tokens per answer than a standard chat model answering the same question. For rough scale, two non-reasoning flagships:

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

Calling o1 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/o1",
    "messages": [{"role": "user", "content": "Work through this scheduling conflict step by step and propose a fix."}],
    "max_tokens": 1200
  }'

If a response comes back empty or truncated, raising max_tokens further is the first thing to try — it is the most common fix with reasoning-family models.

When o1 earns its extra latency and cost

• Multi-step problems where an early mistake ruins the final answer — debugging against a stack trace, working through a scheduling or logistics conflict, planning a sequence of dependent steps.
• Tasks worth paying for the reasoning time, rather than high-frequency calls where a mini or nano-class model would answer fast enough and cheaply enough.
• Not a fit for simple lookups or short-form replies — route that traffic to a smaller model and reserve o1 for the subset that genuinely needs deep, step-by-step reasoning.

Frequently asked questions

Is o1 available in India through unoblox? Yes, openai/o1 is callable on https://api.unoblox.ai/v1 right after signup, billed in rupees against your wallet balance.

What does o1 cost on unoblox? It isn't in our fixed indicative price list; check the live ₹ rate on its model page before estimating a workload's cost.

Why does my o1 request need a higher max_tokens value? o1 uses part of the completion-token budget for internal reasoning before writing the visible answer, so a low max_tokens can cut the response off. Set it to at least 1000 as a starting point.

Does o1 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 o1 and a non-reasoning model like GPT-4.1 with the same key? Yes, the same ub-gw- key works across every model in the catalogue — only the model field changes.

Is GST input tax credit available on o1 usage? Yes, o1 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