GPT-5 pricing in India (₹)

GPT-5 in rupees: ₹126 input / ₹1,008 output

GPT-5 is live on unoblox at ₹126 per 1M input tokens and ₹1,008 per 1M output tokens — one OpenAI-compatible endpoint, one API key, one monthly rupee invoice. No international card, nothing to reconcile at the end of the month.

What a real request costs

A 10K-token prompt with a 2K-token response:

• Input: (10,000 / 1,000,000) × ₹126 = ₹1.26
• Output: (2,000 / 1,000,000) × ₹1,008 = ₹2.02
• Total: ₹3.28, itemized on your GST invoice with input-tax credit claimable.

Scale that up: 1,000 such requests a day is roughly ₹3,280/day, or about ₹98K/month — predictable, and it doesn't move unless your usage does.

GPT-5 vs GPT-4o vs GPT-5 mini

Model	Input ₹	Output ₹	Best for
GPT-5	126	1,008	Multi-step reasoning, long-form analysis
GPT-4o	252	1,008	Vision, real-time/latency-sensitive apps
GPT-5 mini	25.2	201.6	High-volume, well-defined tasks

GPT-5 costs half of GPT-4o on input while generally reasoning better on multi-step problems — GPT-4o's advantage is native vision and slightly snappier responses for simple turns. GPT-5 mini is the pragmatic default for anything that doesn't need frontier reasoning: it's 5× cheaper on input than GPT-5 and handles the bulk of production chat/classification workloads without a noticeable quality gap.

When to reach for GPT-5 specifically

• Long documents: 128K context window comfortably holds large contracts, codebases, or multi-file reviews.
• Multi-step reasoning: planning, research synthesis, and tasks where the model needs to hold several constraints at once.
• Agentic tool use: chained tool calls where a reasoning misstep early on compounds.

For single-turn classification, short replies, or simple extraction, GPT-5 mini gets the same practical outcome for a fraction of the cost — there's little reason to default to the larger model everywhere.

How to call it

from openai import OpenAI

client = OpenAI(
    base_url="https://api.unoblox.ai/v1",
    api_key="ub-gw-…"
)

response = client.chat.completions.create(
    model="openai/gpt-5",
    messages=[{"role": "user", "content": "Your query"}],
    max_tokens=2048
)

Streaming (stream=True), tool calling, and JSON mode all work exactly as they do against OpenAI directly — only the base URL and key change.

Frequently asked questions

How much context does GPT-5 support? 128K tokens — enough for long documents, multi-turn agent sessions, or retrieval-augmented prompts with substantial context.

Do I pay for failed or empty requests? No. A 0-token response (a parse error, a moderation block, a client-side timeout) is never billed — only completed responses with real output tokens.

Can I stream GPT-5 responses? Yes, set stream=True in the OpenAI SDK exactly as you would against OpenAI's own endpoint.

Is GPT-5 fine-tunable through unoblox? Not yet. Few-shot prompting covers most customization needs; the SDK's caching options help reduce repeated-context cost.

Should I default to GPT-5 or GPT-5 mini? Start with GPT-5 mini and only move up if you can show GPT-5 measurably improves your actual output — for most production traffic, mini is indistinguishable in outcome and meaningfully cheaper.

Is pricing the same for every request size, or does it change with volume? The per-token rate is flat regardless of volume — there's no separate enterprise discount tier for GPT-5 today. What changes with volume is your total bill, not the rate.

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