Deep Cogito, a new AI startup founded by Drishan Arora and Dhruv Malhotra, has launched from stealth with a family of hybrid models called Cogito 1 that allow users to toggle between fast responses and advanced reasoning.
Built on open-source foundations like Meta’s Llama and Alibaba’s Qwen, Cogito 1 models — ranging from 3B to 70B parameters — combine performance with adaptability. The largest model reportedly surpasses DeepSeek R1 and Llama 4 Scout in key benchmarks.
Malhotra, formerly of DeepMind, and Arora, previously at Google, said the models were trained in just 75 days and signal a step toward building “general superintelligence.” All models are now accessible via Fireworks AI and Together AI.
