SAN FRANCISCO, Calif.(Diya TV) — Mirendil, an artificial intelligence startup founded by former Anthropic researchers, has raised $1 billion in Series A funding. The investment marks a major step for the company as it works to change how researchers develop and train advanced AI models.
Kleiner Perkins and Clearlake Capital led the funding round. Several other investors also joined, including Battery Ventures, Doerr Capital, Google, M13, Pear VC and S32.
The San Francisco-based startup aims to automate large parts of artificial intelligence research and development. Its technology could help smaller teams build powerful AI models without the large research departments and budgets that major technology companies typically require.
The latest investment also highlights growing interest in AI systems that can perform complex research tasks with limited human involvement.
Today, developing advanced AI models requires significant computing power, money, and technical expertise. Major AI companies often depend on large teams of researchers and engineers to train models, test new ideas and improve performance.
However, Mirendil wants to change that approach.
The company is developing AI agents that can handle many stages of the research process. These systems can write code, conduct experiments, and support model development. As a result, smaller research teams could potentially complete projects that once required much larger organizations.
Moreover, Mirendil wants its technology to reduce the time and cost involved in building advanced AI systems.
The company’s broader goal is to make AI research more efficient. It believes automated research tools could allow individual scientists and smaller teams to compete in an industry dominated by well-funded technology companies.
Mirendil has shared an example of how its technology works. According to the company, one researcher used its AI agents to train a coding model called Mirendil-Base from scratch.
The project took just 3.5 months to complete. It also required less than $10 million in computing resources.
Mirendil claims its AI agents generated more than 99% of the code used throughout the process. The company presented the project as evidence that AI agents can handle a significant share of complex model-development work.
The achievement could point to a different approach to AI development. Traditionally, researchers write and manage much of the code needed to train models. They also run experiments, review results, and make adjustments throughout the process.
However, AI agents could take over many of these responsibilities. This could allow researchers to focus more on setting goals, reviewing results, and making important technical decisions.
Still, the company’s claims highlight its own reported results. Independent verification would help establish how well the approach performs across different research projects.
In addition to its AI research agents, Mirendil has developed two other products: Fern and Lor.
Fern is a model that supports a context window of up to 10 million tokens. A context window determines how much information an AI model can process during an interaction.
A larger context window can help a model work with extensive documents, research materials and code. Therefore, Fern could support projects that require AI systems to handle large amounts of information at once.
Meanwhile, Lor serves as a platform for coordinating AI agents. It allows hundreds of agents to write code, run experiments and work on research tasks over extended periods.
Together, these products form part of Mirendil’s broader effort to automate AI development. The company aims to build systems that can manage complex research workflows rather than simply respond to individual prompts.
The new $1 billion investment comes less than six months after Mirendil raised $200 million in seed funding.
The rapid fundraising highlights investor interest in companies developing AI systems that can automate research and technical work.
Kleiner Perkins and Clearlake Capital led the latest round. Other participants included Battery Ventures, Doerr Capital, Google, M13, Pear VC and S32.
However, Mirendil has not disclosed the valuation attached to its completed Series A round.
Bloomberg previously reported that the startup was seeking a $5 billion valuation. That figure reflected reported fundraising discussions, not a confirmed valuation for the latest investment.
Mirendil’s funding arrives as AI companies compete to develop increasingly capable models. At the same time, the cost and complexity of AI research remain major challenges for developers.
If the company’s technology delivers on its goals, AI agents could help smaller teams conduct research with fewer resources. This could also change how companies approach software development, model training and scientific experiments.
Nevertheless, automated AI research still faces important questions. Researchers will need to assess the reliability, accuracy and safety of systems that make technical decisions and run experiments with limited supervision.
For now, Mirendil’s $1 billion Series A round gives the startup significant financial support as it develops its technology. Its next challenge will be to demonstrate that AI agents can deliver consistent results across a wider range of research tasks.