San Francisco-based Mirendil said Thursday it raised $1 billion in a Series A to develop frontier AI research tools for scientists and engineers, addressing what it says is limited access to such capabilities.
Kleiner Perkins and Clearlake Capital led the round. The company also listed S32, Google, Pear VC, M13, Gil & Co, Doerr Capital and Battery Ventures among the investors. It did not disclose a valuation, individual investment amounts or a detailed spending plan.
Mirendil says it wants to extend frontier AI tools to researchers beyond the labs able to build them. The company points to drug discovery, chemistry, biology and robotics, arguing that the cost of developing AI capacity leaves it in the hands of relatively few labs.
Its Lor platform remains in private beta. Mirendil says users can delegate coding and experimental work to groups of hundreds of agents, with tasks extending across several days. Access is by invitation and free during the beta.
The company says it developed its Fern model by post-training Mirendil-Base, its coding model. Mirendil describes Mirendil-Base as having between 100 billion and 500 billion parameters and says Fern supports a context window of up to 10 million tokens. Fern access is listed as coming soon, and the company has not posted general-release dates for either product.
Mirendil said one researcher and AI agents trained Mirendil-Base from scratch in 3.5 months, with compute costs below $10 million. The company said agents wrote more than 99% of the code, ran experiments and diagnosed failures; the researcher set the research direction and judged the results. EconoTimes previously covered Mirendil’s account of that AI-assisted training process.
The Series A follows a $200 million seed round announced June 24, led by Andreessen Horowitz and Kleiner Perkins, with NVIDIA among the investors named by Mirendil.
