Deep Cogito Raises $43M to Build AI Models Enterprises Can Own

Deep Cogito has raised a $43 million Series A to build specialized AI models that enterprises can train on proprietary data and own. The financing backs a different enterprise bargain: customers can shape a model around internal work and retain the resulting system.

TQ Ventures led the round, joined by Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons and Zscaler. Deep Cogito did not disclose its Series A valuation.

$43 millionSeries A

Deep Cogito announced the Series A on August 26.

More than $56 millionTotal outside funding

The new round brings the company’s outside funding above $56 million.

The sale starts after pretraining

Deep Cogito operates a platform for enterprises to build custom models from internal data alongside research on self-improving models. Its Cogito family is open source, and the latest cited release is Cogito v2.1 671B.

Its approach begins with open pretrained checkpoints, then applies post-training to improve reasoning, tool use and task-specific performance. Deep Cogito calls its method Iterated Distillation and Amplification, or IDA: a model uses added computation to seek a stronger answer, and that reasoning process is then distilled into the model’s parameters.

Zscaler is a customer signal, not proof of scale

Zscaler provides a concrete enterprise connection for the pitch. The cybersecurity company named Deep Cogito a technology alliance partner for Project AI-Guardian in June, and it is both an investor and a user of Deep Cogito’s enterprise platform. The alliance involved planned integrations, not a disclosed customer deployment; the scale and outcome of Zscaler’s use have not been disclosed.

The money moves the test from models to deployments

Deep Cogito plans to use the funding to expand its enterprise business, add research staff and release more open-source models. Founded in 2024 by former Google employees Drishan Arora and Dhruv Malrana, it previously raised a reported $13 million seed round led by Benchmark.

The commercial test is whether models tailored to a company’s data and work can outperform general models on tasks customers will pay to improve. Deep Cogito has released open models; the new funding is intended to make enterprise customization a business that can scale.