Deep Cogito

About Deep Cogito

Deep Cogito is a San Francisco AI research lab focused on post-training, large-scale reinforcement learning and recursive self-improvement. It builds the Cogito family of open-weight models, released across sizes from 3B to more than 600B parameters, and applies the same post-training engine to create specialised models trained on enterprises' proprietary data, decisions and outcomes. A core research direction is Iterated Distillation and Amplification (IDA), in which a model uses extra computation to produce better answers and then distils those improvements back into its weights, with the long-term aim of models that progressively improve beyond the limits of human-generated training data. The company was founded in 2024 by CEO Drishan Arora, who led Gemini post-training for Google AI Search, and Dhruv Malrana, who led the AI Search product from inception, including AI Mode and AI Overviews. After a $13 million seed round backed by Benchmark and South Park Commons, on 26 August 2026 Deep Cogito announced a $43 million Series A led by TQ Ventures, with Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons and customer and strategic investor Zscaler, bringing total funding to more than $56 million, to grow its research and engineering team, scale training infrastructure and expand enterprise work. Sources: Business Wire 26 Aug 2026; deepcogito.com.