In Sanskrit “Krutrim” means artificial, and hence the name.
Krutrim has been trained on 2 trillion tokens and unique datasets.
There will be a larger more complex model called Krutrim Pro that will be launched in the next quarter.
AI is the most influential and trending thing currently in the tech world. And it is affecting all of us daily. Currently, OpenAI’s ChatGPT rules over the market. However, all major tech companies are fighting against time to bring out a Generative AI model to compete with Chat GPT. The craze has traveled to India too. Ola co-founder Bhavish Aggarwal’s artificial intelligence venture Krutrim Si Designs has now unveiled Krutrim, which will join the AI race and is the first of the multilingual AI models that the company has to offer. This is also India’s very first AI multimodal model.
In Sanskrit “Krutrim” means artificial, and hence the name. Aggarwal’s AI venture, the Krutrim Si Designs was inaugurated in April 2023. It was set up in partnership with Krishnamurthy Venugopala Tenneti, who is one of the board members of ANI Technologies Ltd, which owns Ola Cabs and Ola Electric.
Let us talk about its AI model now. Krutrim is being called India’s very own AI. This AI model will be offered in two sizes: a base model called Krutrim and Kurtrim Pro. Krutrim has been trained on 2 trillion tokens and unique datasets. Along with this, there will be a larger more complex model called Krutrim Pro which is expected to arrive in the early part of next year. It is said to have more sophisticated problem-solving and task-execution capabilities.
Now, what’s the first thing that comes to your mind? Is it better than other AI chatbots such as ChatGPT and Bard? Well, let’s see what Ola has to say about it.
Is Krutrim better than other AI models?
Aggarwal informed us during the launch event that the Krutim-powered AI chatbot is very similar to the interface that you get to see on ChatGPT. He further informed that the AI can understand over 20 languages and on top of that it can even generate 10 Indian languages. The training for this chatbot has been done using 2 trillion tokens. Further, this Indian AI model outperformed GPT-4 and Llama 2 in the Indic language, as informed by Ola itself.
Follow Us
Mustafa Khan
Mustafa is a young tech journalist who tells it like it is, cutting through buzzwords to deliver straightforward smartphone reviews. He’s the office go-to for insider tips and quick demos, and his video content doesn’t waste anyone’s time. When he’s off the clock, he geeks out over cars, photography, and hunting down the best spot for Indian food. View Full Profile
