<ul><li>Startup Deep Cogito Inc. launched with language models that claim to outperform open-source alternatives.</li><li>The Cogito v1 series models are available in sizes ranging from 3 billion to 70 billion parameters.</li><li>Deep Cogito uses a hybrid architecture combining large language models and reasoning models.</li><li>The company's IDA training method improves the models' performance by distilling knowledge from the generated answers.</li></ul>

Deep Cogito releases open-source language models that outperform Llama

Discover more