Skip to main content

Mirendil announces $1 billion Series A for AI research tools

13 NOVEMBER 2026·2 MIN READ·5 SOURCES

Mirendil announced a $1 billion Series A on November 12, 2026, to support its work making frontier AI research and development more accessible. Kleiner Perkins and Clearlake Capital led the round.

Mirendil announces $1 billion Series A for AI research tools

Key takeaways · 4

  • 01

    The Series A was $1 billion, led by Kleiner Perkins and Clearlake Capital.

  • 02

    Mirendil says the Mirendil-Base training experiment used less than $10 million in compute, including agent token spending and data processing.

  • 03

    Mirendil says Fern supports a context window of up to 10 million tokens and was designed for high-volume agentic work.

  • 04

    Lor was free during its invitation-only private beta and available on macOS, Linux and the web.

Funding for broader access

Mirendil announced a $1 billion Series A on November 12, 2026, with Kleiner Perkins and Clearlake Capital leading the round.[1][2] Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 also participated, along with other new and existing investors.[2] Mirendil says it is building tools to make frontier AI research and development accessible to scientists and engineers.[2] The company says the funding will help it pursue work that previously required a team or an institution with fewer people.[2]

A model-training example

To illustrate its approach, Mirendil says one researcher working with agents pretrained Mirendil-Base from scratch in 3.5 months, with less than $10 million in compute.[2] It says AI agents wrote more than 99% of the code for the experiment.[3] The reported compute figure included agent token spending and data processing, according to Mirendil.[3] The company says all training data came from GitHub and other publicly available code, web pages and datasets.[3] Mirendil-Base is in the 100-billion-to-500-billion-parameter range.[3]

Fern targets long-context work

Mirendil says Fern is its first model and was pretrained from scratch.[4] Fern supports a context window of up to 10 million tokens, which the company says can hold about half a million lines of code.[4] Mirendil designed Fern for high-volume agentic tasks such as debugging training runs, analyzing agent traces and auditing training data.[4] The company listed access to Fern as coming soon.[4] The stated use cases and availability make the model’s intended role clearer, but the evidence does not provide pricing or access dates.

Lor and earlier company announcements

Mirendil says its Lor agents can split a job across dozens of agents, with hundreds running across a user’s projects, and can unblock themselves when something breaks.[5] The company also says Lor agents can request, configure and release compute nodes themselves.[5] Lor was available by invitation only in a private beta, free during that beta, and supported macOS, Linux and the web.[5] Windows, iOS and Android support were listed as coming soon.[5] Earlier, Mirendil disclosed a $200 million seed round on June 24, 2026, and announced a multi-year Google Cloud partnership on August 6, 2026.[1]

Mirendil’s announcement pairs a large funding round with examples of research and development it says a small team can carry out using agents. Professionals evaluating agent-based workflows can use the details—and the stated beta and access limits—to distinguish company-reported demonstrations from tools they can access today.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 13 November 2026

    Mirendil announces $1 billion Series A for AI research tools

Sources

AI fluency, one session a day, built for your work.