
Mirendil raises $1 billion Series A for AI research and model development
The AMW Read
The reported $1 billion Series A materially expands funding for a known player, while its automation claims remain unverified.
Mirendil raises $1 billion Series A for AI research and model development
Mirendil has announced a $1 billion Series A co-led by Kleiner Perkins and Clearlake Capital, with participation from Google, according to Brief IA. The company plans to use its algorithms to develop more capable language models. Its products include Lor, which automates tasks in model training, and Fern, a coding model built on Mirendil-Base. Mirendil says Fern can process 10 million tokens per prompt.
The round extends the company's push to automate AI research, which AI Market Watch covered when Mirendil raised a $200 million seed round. Its approach combines foundation-model development with agents that run and compare training experiments. Mirendil says hundreds of agents divide the work and that Lor corrects errors detected during evaluations. The company also claims agents wrote more than 99% of Mirendil-Base's code and handled infrastructure, data collection, and parts of training design. Those claims are company-reported; the article does not provide independent verification.
For investors and model builders, the question is whether those reported automation gains hold across projects and reduce the full cost of development. Mirendil says one researcher used its automation to build a coding model for $10 million over roughly three and a half months. The article does not establish how that model compares with rivals or disclose a breakdown of the reported cost.
