Best Cheap LLM API in India: ₹ Price List

If you're building in India and every rupee of inference cost matters, the good news is unoblox's catalog has real range — from a completely free model to sub-₹15 tiers that are cheap enough to run at high volume without a second thought. Here's the actual ₹ price list, cheapest first.

The cheapest models on unoblox today, ranked

Model	Model id	Input (₹ / 1M tokens)	Output (₹ / 1M tokens)
Qwen3 1.7B	qwen/qwen3-1.7b	₹0 (free)	₹0 (free)
Qwen3 235B-A22B	qwen/qwen3-235b-a22b-instruct-2507	₹9.07	₹55.44
DeepSeek V4 Flash	deepseek-ai/deepseek-v4-flash	₹9.07	₹18.14
Llama 4 Scout	meta-llama/llama-4-scout-17b-16e-instruct	₹10.08	₹30.24
Gemma 4 31B	google/gemma-4-31b-it	₹13.10	₹38.30
Qwen3.8-27B	qwen/qwen3.8-27b	₹16.32	₹48.96
GPT-5 mini	openai/gpt-5-mini	₹25.2	₹201.6

Qwen3 1.7B costs nothing to call — a genuinely free tier, not a time-boxed trial. Among paid models, DeepSeek V4 Flash has the lowest output price in this table at ₹18.14 per million tokens, while Qwen3 235B-A22B matches its input price exactly at ₹9.07.

Why "cheap" still needs a fit check

A low ₹ rate only saves money if the model is actually good enough for your task — routing production traffic to a model that produces unusable output and needs a retry, or a fallback to a pricier model, can end up costing more than picking the right model the first time. Treat this list as a shortlist to test, not a guarantee.

Where each tier tends to fit

• Free tier (Qwen3 1.7B): prototyping, internal tools, low-stakes classification, or anywhere you want zero marginal cost while you validate an idea.
• Sub-₹15 tier (Qwen3 235B-A22B, DeepSeek V4 Flash, Llama 4 Scout): high-volume production workloads — support triage, tagging, extraction — where per-call cost needs to stay low.
• ₹15 to ₹30 tier (Gemma 4 31B, Qwen3.8-27B): slightly heavier tasks where you want more headroom, including vision input on Qwen3.8-27B.
• GPT-5 mini: a budget option if you specifically want OpenAI's ecosystem and tuning at a lower price than the GPT-5 flagship.

Calling the cheapest tier from your existing code

curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen/qwen3-1.7b","messages":[{"role":"user","content":"Tag this ticket as billing, technical, or general."}]}'

Every model above uses the identical request shape — only the model id changes, and your GST invoice covers all of them together.

Frequently asked questions

Is Qwen3 1.7B really free, with no catch? Yes — it's priced at ₹0 for both input and output tokens on unoblox; there's no separate trial period tied to it.

What's the cheapest paid model for high-output workloads? DeepSeek V4 Flash, at ₹18.14 per million output tokens, is the lowest output rate among the paid models listed here.

Will a cheap model be good enough for my use case? It depends entirely on the task — test it against your own prompts before routing production traffic, since ₹ cost and output quality are separate questions.

Are there hidden fees on top of the per-token ₹ rate? unoblox bills usage on one monthly GST invoice; check your dashboard and the model's live page for the current rate before scaling up.

Can I mix a free model and paid models under one account? Yes — one key and one invoice cover every model in the catalog, free or paid.

Do these cheap models support the same features as premium ones, like streaming and tool calls? Yes, all models are served through the same OpenAI-compatible endpoint with consistent streaming and tool-calling support.

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