Claude API in India

Claude Models Available in India — Rupee Billing, Indian Entity

Anthropic's Claude Opus and Claude Sonnet are routable through unoblox's single OpenAI-compatible gateway from an Indian-registered entity. Build reasoning-heavy applications with full local billing: monthly rupee invoices, GST input tax credit, and zero currency friction.

Pricing & model selection

Model	Input (₹/1M)	Output (₹/1M)	Context	Best for
Claude Opus	504	2,520	1M tokens	deep analysis, research
Claude Sonnet	201.6	1,008	1M tokens	production apps, speed

Live rates are always visible on the /models dashboard.

Set up in two steps

from openai import OpenAI

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

response = client.chat.completions.create(
    model="anthropic/claude-sonnet-5",   # or anthropic/claude-opus-5
    messages=[{"role": "user", "content": "Draft a GST-compliant invoice checklist"}]
)
print(response.choices[0].message.content)

Streaming, tool calls, vision, and JSON mode work exactly as they do against Anthropic's own API.

Why Claude through unoblox

₹-native pricing, no Anthropic account needed: sign up once, get an Indian GST invoice, and skip international payment setup entirely.

Extended thinking on Opus: Claude Opus supports step-by-step extended reasoning (thinking: {"type": "enabled", "budget_tokens": 10000}) — useful for code review, research synthesis, and multi-turn agent planning.

One key, every model: the same ub-gw-… key also routes to GPT, Qwen, DeepSeek, and more — swap the model field, not your integration.

GST compliance included: one monthly invoice for your books, input tax credit claimable by your finance team.

Claude vs other unoblox models

Model	₹/1M in / out	Where it fits
Claude Opus	504 / 2,520	hardest reasoning, research, agent planning
Claude Sonnet	201.6 / 1,008	production APIs, chat, everyday coding
GPT-4o	252 / 1,008	general-purpose, strong tool use
DeepSeek V3.2	26.21 / 38.30	budget reasoning at high volume
Qwen3 1.7B	0 / 0	prototyping, low-stakes steps

A common pattern: default to Sonnet, escalate to Opus only when a response looks under-baked, and fall back to DeepSeek or the free Qwen tier for anything routine.

Frequently asked questions

Does unoblox support Claude's vision capabilities? Yes. Both Opus and Sonnet accept image and PDF inputs through the standard messages array.

What's the difference between Opus and Sonnet in practice? Opus is tuned for the hardest reasoning and multi-step problems; Sonnet trades a little of that depth for lower cost per call, which suits most production APIs.

How do I migrate from Anthropic's direct API? Change two lines — base_url and the API key — and swap the model id to anthropic/claude-opus-5 or anthropic/claude-sonnet-5. Nothing else in your code changes.

Is there a free tier or trial? New accounts receive trial credits, visible on your dashboard right after signup.

What if I hit rate limits? Request a higher tier from your account settings, or contact support for volume pricing.

Can I use Claude alongside open models in the same app? Yes — route hard reasoning to Claude and routine calls to a cheaper model like Qwen3 1.7B (free) or DeepSeek, using the same key throughout.

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