Tamil NLP API in India | unoblox

Building a Tamil-language feature — support-ticket replies, document summarisation, translation — usually means picking a strong general-purpose multilingual model rather than waiting for a Tamil-specific one. unoblox puts several such models behind one OpenAI-compatible endpoint, billed per token, in rupees.

What "Tamil NLP" covers in production

In practice this spans a handful of concrete jobs: triaging and drafting replies to Tamil support tickets, summarising Tamil documents or transcripts, generating Tamil marketing or product copy, translating between Tamil and English, and light transliteration help. None of these need a model trained only on Tamil — they need a capable general-purpose model prompted correctly.

General-purpose models, not a Tamil-specific one

To be direct about it: unoblox doesn't offer a Tamil-specific fine-tuned model. What it offers is a set of general multilingual models — from the Qwen3, Llama 4, DeepSeek and GPT families — that are commonly used for Indian-language text tasks including Tamil. That's a qualitative statement about how these model families are used, not a claim of any specific benchmarked accuracy for Tamil.

Multilingual models available today, in rupees

Model	Input ₹ / 1M tokens	Output ₹ / 1M tokens
Qwen3 235B-A22B	₹9.07	₹55.44
Llama 4 Maverick	₹20.16	₹80.64
DeepSeek V4 Flash	₹9.07	₹18.14
GPT-5	₹126	₹1,008
Qwen3 1.7B	Free (₹0)	Free (₹0)

Prompting and billing notes for Tamil text

• Scripts like Tamil can tokenize into more tokens per sentence than the equivalent English text, depending on the model's tokenizer — factor that into cost estimates rather than assuming a 1:1 ratio with an English example.
• State the output language explicitly in the prompt or system message. Models generally default to matching the input language, but an explicit instruction is more reliable than relying on that default.

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-ai/deepseek-v4-flash","messages":[{"role":"user","content":"Reply to this customer question in Tamil: When will my order arrive?"}]}'

Frequently asked questions

Is there a Tamil-specific model on unoblox? No. The catalog is general-purpose models — see /models for the current list.

Does Tamil text cost more per request than English? Billing is per token either way, but Tamil script can tokenize into more tokens for the same idea, which can raise the per-request cost compared with an equivalent English prompt — a tokenizer property, not a Tamil-specific surcharge.

Which model should I start with for a Tamil chatbot? Try a couple from the table above through /models — there's no single verified best answer to state here.

Is there a free way to prototype? Yes, Qwen3 1.7B is free (₹0).

Is my Tamil text processed in India? Depends on the model. Only select unoblox-hosted small models on unoblox run on India infrastructure; most catalog models don't guarantee residency — check /models.

Do I need a different SDK to send Tamil text? No — plain UTF-8 text over the same OpenAI-compatible endpoint as any other request.

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