AI & MLNews

SquadStack Introduces Maya AI Model for Sales Conversations, Reports 96.7% Benchmark Pass Rate

Vikas Gulati Co-Founder & CTO

SquadStack has launched Maya, a large language model (LLM) developed specifically for sales conversations. The company claims the model outperformed 10 AI models from developers including OpenAI, Google, DeepSeek and Alibaba in its benchmark evaluation using real Indian sales calls.

According to SquadStack, Maya achieved a 96.7% pass rate across 9,962 conversational exchanges drawn from 471 sales calls. The benchmark assessed models across seven parameters, including conversational context retention, adherence to campaign objectives and correct tool usage.

In the company’s reported comparison, Google’s Gemma 4 31B achieved a 92.5% score, followed by DeepSeek V4.1 Flash at 92.2%. OpenAI’s GPT-6 Luna recorded 89%.

Maya Designed for Real-Time Sales Conversations

SquadStack said Maya recorded a median response time of 278 milliseconds on its production endpoint in India. The model was trained using more than 40 crore sales interactions collected over six years. These conversations were tagged according to outcomes such as completed sales, callbacks and customer drop-offs.

The company said Maya currently processes more than 50 lakh conversations daily in Hinglish and other Indian languages, handling over one trillion tokens every month.

“A live sales call is the hardest place to run an LLM,” said Vikas Gulati, Co-founder and CTO of SquadStack. “Thousands of calls can start in the same second, and the customer hears every slow turn. So we built Maya for this one job, with the sales behaviour in its weights rather than in a long prompt. Maya has already processed more than 5 trillion tokens, and it now handles over a trillion tokens a month.”

Second AI Model After Arth V2

Maya is SquadStack’s second AI model, following the launch of its speech recognition model, Arth V2. The new LLM is deployed within the company’s Voice AI agents for enterprise customers.

Unlike general-purpose language models, Maya has been developed specifically for sales interactions, with a focus on conversational context, campaign objectives and tool execution. Its training approach uses historical sales conversations and their outcomes to support performance in real-world calling environments.

India-Hosted Model with Continuous Learning

SquadStack said Maya is hosted in India. Its learning system, called Lift, uses outcomes from live calls to identify performance gaps and improve the model over time.

The launch reflects the growing development of specialised AI models designed for specific business workflows, particularly in areas such as sales automation and customer engagement. Through Maya, SquadStack aims to bring domain-specific language model capabilities to enterprise sales conversations across Indian languages.

Related posts

Cloud Box Technologies Secures Titanium-Level Recognition as a Solution Provider with Dell Technologies

admin

Allana Group Appointed Navin Nathani as Group Chief Information Officer

admin

Cloudflare Q3 2025 Report Flags Surge in Hyper-Volumetric DDoS Attacks Driven by Aisuru

admin
×