Tech Mahindra has launched an upgraded version of its large language model (LLM), Project Indus 2.0, which was built with under $5,00,000. The company has also announced plans to expand the model’s scope to include Bengali audiences and develop an agentic application in Hindi and its dialects. An agentic application is a software system that can independently work towards goals for a user, often driven by AI such as LLMs.
The company has reported that the LLM has witnessed 8,000 downloads so far and is expanding its database and adding more data for dialects through supervised fine-tuning. Tech Mahindra is also working on a project to build a small model with 1.2 billion parameters, which would make it ideal for deployment on chips. This would be useful in edge scenarios such as farms, factories, and the Indian automotive industry.
An independent study conducted by Assam Kaziranga University compared various models and demonstrated that Indus outperformed many of its counterparts in tokenisation efficiency. The study found that the speed of tokenisation is influenced by the number of parameters and the model’s size, and that Indus has shown it is possible to outperform even larger models by using optimised techniques.
Tech Mahindra is leveraging LLMs to enhance its services for customers and improve delivery. The company is using OpenAI models to fine-tune its LLMs, and the usage varies depending on the specific use case, which is determined through research to compare the models and select the appropriate one. The Indus LLM has many applications across India, including supporting low-income families with financial services in their dialects and providing chat-based AI features to students.