Mirendil raises $1bn Series A for agents that run AI research

Mirendil raises $1bn Series A for agents that run AI research

12 November 2026: Mirendil says it has raised $1 billion in a Series A led by Kleiner Perkins and Clearlake Capital, with Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 also taking part. The company builds tools that let agents do the hands-on work of AI research, and it argues that once agents do that work, team size stops limiting what a lab can build.

Not yet independently verified. Mirendil's release gives no valuation, headquarters, founders or founding year; those appear only in press reports, so we leave them out. We will update this when it can be confirmed, and remove this note.

One researcher, one model, under $10 million

The company's evidence is its own. It says one researcher working with agents pretrained Mirendil-Base, a competitive coding model, from scratch in three and a half months using less than $10 million in compute. It then post-trained that model into Fern, which it says holds up to 10 million tokens of context, "enough for days of an agent's work."

Those results were published over the three weeks before the round was announced. Mirendil does not give benchmark scores for either model in the funding post, and we have not seen independent evaluations, so the claim that Mirendil-Base is competitive rests on the company's word for now.

Lor, the product buyers can ask for

The part of Mirendil an outside team can actually use is Lor, in private beta. The company describes it as "hundreds of agents writing code and running experiments for days at a time," and says it runs its own work on Lor every day. There is no published pricing and no public sign-up; Mirendil asks teams that want to build with it to write to its contact address.

For research groups and engineering teams, the pitch is a direct one: hand long-running experimental work to a fleet of agents rather than to more staff. How Lor handles compute costs, data access and review of agent output is not described in the announcement.

A very large first priced round

A billion-dollar Series A is unusual, and Mirendil says little about itself beyond the products. "We're still a small team with a singular focus," the post says. It does not state a valuation, headquarters, founding date or founders. Webrazzi reports the founders' names and a 2026 founding date, but we have not found those on the company's own pages.

Agents and models in our catalogue

Our AI model category lists 159 models and our artificial intelligence companies category 1,113 firms. Agent platforms aimed at research, rather than at customer service or coding assistance, are still a small group within them, which is where Lor would sit if Mirendil opens it more widely.

Company profile on TrustList: Mirendil

Sources

• Mirendil: Mirendil Raises $1B to Democratize Frontier AI R&D, 12 November 2026
• Webrazzi: Mirendil funding report, 13 November 2026