Mirendil, an AI lab building models and software to automate parts of AI research, is in advanced talks to raise up to $1 billion at a $5 billion post-money valuation, Briefs Finance reported, citing people with knowledge of the discussions. Kleiner Perkins is in line to lead, and Andreessen Horowitz is considering participation. The financing has not closed.
Rather than offer only a general-purpose model, Mirendil wants to combine specialized models with a research platform. In practice, the proposed system would move work from experimental design and code execution through evaluating results and selecting the next test. Andreessen Horowitz expects engineers and AI researchers to be the initial users, with scientists and other specialists targeted later.
The talks precede Mirendil’s first model release. Briefs Finance said the money would support that launch and further training, with an engineering- and research-focused frontier model planned for early 2027. Mirendil’s website frames the longer-term goal as making frontier AI research tools accessible to labs working in areas such as drug discovery, chemistry, biology and robotics.
The proposed terms would mark a rapid repricing. Mirendil announced a $200 million seed round led by Andreessen Horowitz and Kleiner Perkins on June 24, and Briefs Finance said that financing valued the company at $1 billion. The new post-money figure would be five times the earlier valuation only three months later, before the planned first model ships.
Mirendil launched with 20 researchers and engineers from Anthropic, xAI, Google DeepMind and OpenAI. Andreessen Horowitz said co-founder Harsh Mehta built the first version of Anthropic’s internal autoresearch platform. That background is central to Mirendil’s pitch: take a research loop already used inside a frontier lab and turn it into a system engineers and researchers outside those labs can use.