Market · AI in practice
Mirendil raises US$ 1 billion to bring frontier AI research to small teams
Mirendil announced a US$ 1 billion Series A led by Kleiner Perkins and Clearlake Capital, for tools that let scientists and engineers do frontier AI research with agents.
What the sources report
Mirendil raised US$ 1 billion in a Series A led by Kleiner Perkins and Clearlake Capital, with participation from Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32, among other new and existing investors. The announcement came out on the company's website and on PR Newswire and does not disclose a valuation.
The company says it wants to put frontier AI research and development within reach of every scientist and engineer. In the first results it shared before the round, Mirendil says a single researcher, working with agents, pretrained Mirendil-Base, a competitive coding model, from scratch. The result is the company's own.
After pretraining, Mirendil-Base was post-trained into Fern, which holds up to 10 million tokens of context, enough, according to the company, for days of an agent's work. Lor, in private beta, puts hundreds of agents to work writing code and running experiments for days at a time, and Mirendil says it runs on Lor every day. The company describes itself as a small team and says the round is meant to reach a point where one person can take on what used to need a team, and a team what used to need an institution.
My take
In banks and fintechs, building a proprietary model usually runs into the size of the data team. If one researcher with agents can run a pretraining, as Mirendil says, the bottleneck for a credit, fraud or customer service model moves from headcount to the quality of the data and control over what the agents do.