Text summarization API in India (high-volume, ₹ pricing)

Text summarization API in India

Summarization is a workhorse task: news feeds, research abstracts, support tickets, content review. On unoblox, pick your model (Qwen for budget, Claude for quality) and pay in ₹ rupees. One API key, one invoice.

Here's the 3-step setup + real ₹ costs.

Why unoblox for summarization

• Cost: Qwen3 at ₹26/1M tokens (10× cheaper than Claude)
• Speed: Summarize 100 docs/minute at no additional cost
• Quality: DeepSeek V3.2 and Claude preserve key facts perfectly
• Batch friendly: Summarize 1000s of docs overnight
• ₹ billing: GST invoice, input-tax-credit claimable

Recommended models

• Qwen3 Max (₹120.95 input / ₹604.77 output): Best quality-to-cost for high-volume
• DeepSeek V3.2 (₹26.21 input / ₹38.30 output): Cheapest, excellent for factual summaries
• Claude Sonnet (₹201.6 input / ₹1,008 output): Best for preserving tone and nuance

Step 1: Set up the client

from openai import OpenAI

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

Step 2: Summarize a single document

article_text = """Your long article/research paper here..."""

response = client.chat.completions.create(
    model="deepseek-ai/deepseek-v3.2",
    messages=[
        {"role": "system", "content": "Summarize in 3 bullets. Keep facts, drop opinion."},
        {"role": "user", "content": article_text}
    ]
)

summary = response.choices[0].message.content
print(summary)

Step 3: Batch summarize 1000s of documents

import time

articles = [...]  # Your 1000+ articles
summaries = []

for article in articles:
    response = client.chat.completions.create(
        model="deepseek-ai/deepseek-v3.2",
        messages=[
            {"role": "system", "content": "Summarize in 3 bullets."},
            {"role": "user", "content": article}
        ]
    )
    summaries.append(response.choices[0].message.content)
    time.sleep(0.1)  # Respect rate limits

Real ₹ cost: 10,000 news articles

Setup:

• 10,000 articles × 500 tokens avg = 5M tokens input
• Summary: ~100 tokens × 10,000 = 1M tokens output

Using DeepSeek V3.2 (cheapest):

• Input: ₹131.05 | Output: ₹38.30
• Total: ₹169.35 for 10k summaries
• Cost per article: ₹0.017

Using Qwen3 Max (better quality):

• Input: ₹604.75 | Output: ₹604.77
• Total: ₹1,209.52
• Cost per article: ₹0.12

Using Claude Sonnet (editorial quality):

• Input: ₹1,008 | Output: ₹1,008
• Total: ₹2,016
• Cost per article: ₹0.20

DeepSeek is roughly 12× cheaper per summary than Claude Sonnet — good enough for factual, structure-preserving summaries; switch to Claude when tone and nuance matter more than raw cost.

Tips for production

1. Adjust prompt per use-case:

   • News: "Summarize in 1 sentence."
   • Research: "Extract key findings in 5 bullets."
   • Support: "Summarize problem and solution."

2. Control output length: Add max_tokens=150 to limit output

3. Batch overnight: Summarize 1000s overnight; no rate-limit stress

4. Store summaries: Database, vector DB, or S3—your choice

5. Monitor ₹ spend: Dashboard shows real-time usage; set alerts

Frequently asked questions

Q: How long can articles be? Up to 8,000 tokens (~5,000 words). For longer, split into sections and summarize each.

Q: Does quality differ by model? Yes. DeepSeek: factual, accurate. Qwen: balanced. Claude: preserves tone. Test on your data.

Q: Can I summarize in Hindi? Yes. Add to prompt: "Summarize in Hindi." All models support 50+ languages.

Q: Should I use streaming? For single summaries: no. For batches: optional. Streaming adds latency but saves memory.

Q: How do I handle formatting (bullets, tables)? Specify in prompt: "Format as markdown table." LLM respects formatting requests.

Q: Can I filter or rank summaries? Yes. Use Claude for a secondary pass. Rate summaries 1–5; save top-rated ones.

Get started in rupees

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