Aditya Birla Capital Servicing Demo
We tested 9 AI models on 300 loan and insurance servicing messages. gpt-oss-120b, an open model run by Dvar in India, got 97.2% fully right, above a 95% bar, at 8.6× lower cost than Gemini 4 Argon.
97.2%
fully right: above the 95% bar
gpt-oss-120b · Gemini 4 Argon 99.2%
8.6×
lower cost per message
Than Gemini 4 Argon, at list price
100%
fraud messages flagged
153 of 153 runs · Gemini 4 Argon 100%
At 10 lakh messages a month (an assumption): ₹1,39,304 with Gemini 4 Argon vs ₹16,200 with gpt-oss-120b per month.
How to read this demo about 5 minutes
Step 1 · 2 minTry it
Type a customer message, in English or Hinglish, or pick one. Both models reply side by side, with the cost of each.
Open Step 2 · 2 minSee the proofAll 9 models ranked against the 95% bar, and every message with each model's answer.
Open Step 3 · 1 minSee how we'd run itSix steps from your own messages to a model you can trust, including fine-tuning on your data.
OpenWords used in this demo
- Leading paid model
- A model from a large AI company such as Google, Anthropic or OpenAI, used through its paid API.
- Open model run by Dvar in India
- A model whose weights are public, hosted by Dvar in India, so messages stay in India.
- Fully right
- Every part of the answer matches the answer key: each request in the message, the product, and any urgent flag.
- The 95% bar
- The share of messages a model must get fully right to be usable here. Set by us for this demo; your team would set the real one.
- Fine-tuning
- Training an open model further on your own messages, so it gets better at your specific requests.
Why now: Aditya Birla Capital already services 40% of customer email with Gen AI email bots. Source