Deep Cogito Announces $43M Series A to Advance the Post-Training Engine for Frontier Intelligence
aiinsider ·
Deep Cogito, a post-training research lab founded by former Google AI Search leaders Drishan Arora and Dhruv Malrana, announced a $43 million Series A round led by TQ Ventures, with participation from Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons, and Zscaler, bringing its total funding to over $56 million. The company focuses on large-scale reinforcement learning and recursive self-improvement, including research into Iterated Distillation and Amplification (IDA), which aims to let models improve their own capabilities beyond human-generated training data. Its methods have been demonstrated through the open-weight Cogito model family, ranging from 3B to 600B+ parameters, showing that strong models can be enhanced through post-training. The funding will be used to expand research and engineering teams, scale training infrastructure, advance future Cogito releases, and grow enterprise work on specialized models built with proprietary data. Zscaler, initially a customer, also joined the round as a strategic investor, noting that Deep Cogito went beyond lightweight customization to train specialized intelligence directly into models. Investors highlighted the importance of post-training as a key AI layer and praised Deep Cogito's combination of frontier research and real-world deployment.
aiinsider