SquadStack.ai Launches Maya, a Sales-Focused AI Model Built for Indian Conversations

Indian AI startup SquadStack.ai has launched Maya, a specialised Large Language Model (LLM) developed for sales conversations. The company describes it as the world’s first Sales LLM and claims that it outperformed models from OpenAI, Google, DeepSeek, Alibaba and NVIDIA in its internal benchmark.

Maya secured a score of 96.7% on the SquadStack Telesales LLM Benchmark, compared with 92.5% for Google’s Gemma 4 31B, 92.2% for DeepSeek V4.1 Flash and 89% for OpenAI’s GPT-6 Luna.

The model has been developed using more than 40 crore real sales interactions collected over six years of sales operations. Each interaction was tagged based on its outcome, including completed sales, callbacks and customer drop-offs, helping the model learn how conversations progress and how to respond in different situations.

Designed to understand the nuances of Indian customer interactions, Maya can interpret informal responses, hesitation and interruptions across Hinglish and other Indian languages. It currently handles more than 50 lakh conversations every day.

The benchmark used 9,962 conversational moments drawn from 471 real sales calls to assess 11 AI models. The models were evaluated on their ability to generate appropriate next responses, with Anthropic’s Claude Haiku 5.5 serving as an independent evaluator. The assessment covered seven parameters, including maintaining the objective of a conversation and using tools appropriately.

Maya maintained conversational context in 96% of turns, compared with 90% for the next-best model. It also selected an appropriate response 95% of the time, against 87% for its closest competitor.

Maya is the second model developed through SquadStack.ai’s in-house model development platform, following its speech recognition model, Arth V2. The company is also working on further improvements that could help the system interpret pauses, changes in tone and customers’ previous interactions.

Built and hosted in India, Maya is designed to keep customer conversations within the country. This could make the model relevant for sectors such as banking, lending and insurance, where data protection and regulatory requirements are important.

While the reported benchmark results highlight Maya’s performance in sales-focused tasks, the evaluation was developed internally and reflects the specific conversational scenarios selected for the assessment.

Picture of Indian Tech Times

Indian Tech Times

Leave a Reply

Your email address will not be published. Required fields are marked *