Mirendil raises $1.2 billion in under 5 months: reported valuation quintuples, aims to take over AI labs with Agents

Beating AI News Flash: Mirendil, founded by a former Anthropic researcher, announced the completion of a $1 billion Series A funding round, led by Kleiner Perkins and Clearlake Capital, with participation from Google, Battery Ventures, and others. Less than 5 months after completing a $200 million seed round in June this year, cumulative funding has reached $1.2 billion.

The Wall Street Journal, citing people familiar with the matter, reported that this round reached a valuation of $5 billion, 5 times that of the seed round, though the company has not officially disclosed the valuation.

What Mirendil aims to do is use Agents to take on the R&D work of an entire AI lab. Last month, it announced an experiment: one researcher working with an AI Agent pretrained Mirendil-Base from scratch in just 3.5 months, with over 99% of the code written by the Agent and compute costs under $10 million. The company then launched the Fern model, which supports a 10 million Token context.

At the end of October, Mirendil also launched Lor, a multi-Agent R&D platform specifically designed to enable AI Agents to collaborate on model development. It can simultaneously orchestrate hundreds of Agents to write code and run experiments, automatically request and release GPUs, and work continuously for days or even weeks. Mirendil also tested having Lor orchestrate 102 Agents at the same time to set up test environments for 100 separate code changes to PyTorch, averaging 8 minutes per task.

Mirendil hopes to hand this capability to external research teams, so that scientists in fields such as drug research, materials, or robotics can develop specialized models without building large AI teams of their own. In addition to supporting its own Fern, Lor can also call models such as Claude, GPT, Kimi, and DeepSeek. Currently, the platform is open only for invite-only beta testing, which is free during the beta period.