Socket Lab has launched the EKA-1 AI model and two compact Small Language Models, according to Prof. Sashikumar, CEO of ZenteiQ. He described the models as designed to respond to day-to-day activities, questions and queries.

The launch includes two different model architectures. One is a dense model with 2 billion parameters. The other is a 5-billion-parameter Mixture of Experts model, with 1.79 billion active parameters.

Two compact model designs

A dense model uses its full model structure when generating a response. The second model uses a Mixture of Experts approach, in which the stated active-parameter count is lower than its total 5-billion-parameter size.

The two systems were presented as Small Language Models. Their stated focus is handling everyday activities and responding to routine questions and queries, rather than being described only by their total parameter counts.

The launch took place in Delhi, with Prof. Sashikumar identifying the systems as compact models. The available announcement does not provide further information about their release platforms, access arrangements, training data, performance benchmarks or commercial terms.

For users, the immediate relevance is the stated focus on everyday interactions. However, the available details do not establish how the models compare with other systems, what languages they support or what hardware requirements they may have.

Conclusion

Socket Lab's EKA-1 launch includes two compact Small Language Models: a 2-billion-parameter dense model and a 5-billion-parameter Mixture of Experts model with 1.79 billion active parameters. Their stated purpose is to handle day-to-day activities, questions and queries.

Frequently Asked Questions

Q. What has Socket Lab launched?

Socket Lab has launched the EKA-1 AI model and two compact Small Language Models.

Q. How large is the dense model?

The dense model has 2 billion parameters.

Q. What is the size of the Mixture of Experts model?

The Mixture of Experts model has 5 billion parameters, with 1.79 billion active parameters.

Q. What are the models designed to do?

They are described as capable of responding to day-to-day activities, questions and queries.

Q. Where was the launch discussed?

The launch was discussed in Delhi.

Q. Who described the models?

Prof. Sashikumar, CEO of ZenteiQ, described the two models as compact models.