OpenAI o3-mini API in India

o3-mini is OpenAI's smaller, faster reasoning model — built for the same step-by-step internal reasoning as the o-series line, at a lower cost and shorter latency than the full-size reasoning models — and it's callable from India on unoblox's single ₹-billed, OpenAI-compatible endpoint.

The one setting worth getting right: max_tokens

Like every model in the o-series, o3-mini reasons internally before producing its visible answer, and that internal reasoning draws from the same completion-token budget as the final response. Set max_tokens too low and a request can come back empty or cut off mid-answer even though nothing went wrong on the API side. Give o3-mini requests at least 1000 completion tokens as a floor, and more for questions with several dependent steps.

Where o3-mini fits next to o1

o3-mini trades some of the deeper reasoning depth of the full-size o-series models for lower cost and faster turnaround — a reasonable default when a task clearly benefits from step-by-step reasoning but doesn't need the largest reasoning budget available. If a task is under-performing on o3-mini, moving up to openai/o1 is usually the next step to try, rather than moving to a non-reasoning flagship.

₹ pricing for o3-mini

o3-mini doesn't have a fixed indicative rate in this note — see the live number on its model page. For scale, two non-reasoning flagships:

Model	Input (₹ / 1M tokens)	Output (₹ / 1M tokens)
GPT-5 mini	25.2	201.6
GPT-5	126	1008

Calling o3-mini from unoblox

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/o3-mini",
    "messages": [{"role": "user", "content": "Check this SQL query for a logic error and explain the fix."}],
    "max_tokens": 1000
  }'

The request shape is identical to calling OpenAI directly — the base URL, the ub-gw- key and ₹ billing are the only differences.

When o3-mini is worth reaching for

• Coding tasks that need a logic check, not just autocomplete — reviewing a query, tracing a bug, or validating an approach before it ships.
• Structured problem-solving with a handful of steps, where a plain chat model tends to skip a step or jump to a plausible-looking but wrong answer.
• High-frequency reasoning calls where the full cost and latency of openai/o1 isn't justified by the size of the problem.
• Not the right fit for the hardest multi-step problems in a workload — route those to openai/o1 and keep o3-mini for the bulk of everyday reasoning calls.

Frequently asked questions

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

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

Why do I need to raise max_tokens for o3-mini? It spends part of the completion-token budget on internal reasoning before writing the visible answer — a low limit can truncate the response. Start at 1000 and adjust from there.

How does o3-mini compare to o1? o3-mini is the smaller, faster, typically cheaper reasoning model of the two; o1 has more reasoning depth for harder multi-step problems.

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

Is o3-mini usage included in the monthly GST invoice? Yes, combined with every other model's usage into one invoice, with input tax credit claimable.

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