SquadStack.ai launches Maya, claims its Sales LLM outperforms models from OpenAI, Google

ET AI (The Economic Times)
Read full story· 3 min
Indian startup SquadStack.ai launched Maya, a Sales-focused large language model trained on over 40 crore sales calls. Maya outperformed competitors including OpenAI's GPT-6 Luna and Google's Gemma 4 31B on a proprietary sales conversation benchmark.

What happened

SquadStack.ai, an Indian startup, launched Maya, a large language model focused on sales conversations. The company trained Maya using data from over 400 million real sales calls collected over six years. In internal testing, Maya achieved a higher score than models from OpenAI, Google, DeepSeek, Alibaba, and NVIDIA on a proprietary benchmark for sales conversations. Maya is designed to handle Indian sales interactions and is currently used in daily conversations across multiple Indian languages.

Key details

  • Maya scored 96.7% on the SquadStack Telesales LLM Benchmark.
  • Google's Gemma 4 31B scored 92.5%, DeepSeek V4.1 Flash scored 92.2%, and OpenAI's GPT-6 Luna scored 89% on the same benchmark.
  • Maya maintains conversational context on 96% of turns and selects the appropriate response 95% of the time.
  • Maya is built and hosted in India, supporting compliance with India's Digital Personal Data Protection Act.

Why it matters

Banks, lenders, insurers, and other businesses in India that rely on sales calls now have access to a language model trained specifically for Indian sales conversations, which can help them improve customer interactions and meet local data protection rules. Companies using Maya can keep customer data within India, addressing regulatory requirements. Organizations handling large volumes of sales calls may benefit from improved automation and accuracy in their sales processes.


More in Business & Enterprise