Mirendil raises a $1 billion Series A to build agents that do the work of an AI lab
Mirendil, an AI lab started by former Anthropic researchers, raised a $1 billion Series A led by Kleiner Perkins and Clearlake Capital on 12 November 2026.
By Emre A.Published
Tech AI LatestMirendil, an AI lab started by former Anthropic researchers Behnam Neyshabur and Harsh Mehta, announced a $1 billion Series A on 12 November 2026, led by Kleiner Perkins and Clearlake Capital with Google among the investors. It follows a $200 million seed round. Mirendil builds Lor, an agent system in private beta, and says one researcher working with agents pretrained its Mirendil-Base coding model from scratch in 3.5 months for under $10 million in compute. No valuation was disclosed.
Key facts
- Announced: 12 November 2026 (Mirendil, press release via PR Newswire)
- Round: $1 billion Series A, led by Kleiner Perkins and Clearlake Capital
- Other investors: Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32
- Earlier funding: $200 million seed led by Andreessen Horowitz and Kleiner Perkins, then an NVIDIA investment
- Company claim: Mirendil-Base pretrained by one researcher with agents in 3.5 months for under $10 million
- Not disclosed: valuation, Lor pricing, general availability date
What Mirendil announced
Mirendil said on 12 November 2026 that it has raised a $1 billion Series A led by Kleiner Perkins and Clearlake Capital. Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 also invested. The company did not state a valuation in the release. Bloomberg reported in September that Mirendil was in talks to raise up to $1 billion at a $5 billion valuation, a figure the company has not confirmed.
Mirendil describes itself as a frontier lab that trains models built to be good at AI research and development, and redesigns the whole lab around them. Its stated mission is to accelerate science and technology by making frontier AI R&D available to more teams. Before this round it raised a $200 million seed led by Andreessen Horowitz and Kleiner Perkins, followed by an investment from NVIDIA, according to the release. That means the Series A is five times the seed, a few months later.
Lor, Mirendil-Base and Fern
The product is Lor, which Mirendil says it is building "to do the work of an entire AI lab". In the company's description, Lor's agents write code, run experiments and iterate on results for days at a time, on any machine and hundreds at once. Lor is in private beta; access is by invitation through a waitlist and is free during the beta.
The headline claim behind the round is a model. Mirendil says one researcher, working with agents, built Mirendil-Base, a coding model, from scratch in 3.5 months with less than $10 million in compute, a figure the company says covers experiments, the final training run and agent token spending. It says agents wrote more than 99% of the code. Mirendil then post-trained the base model into Fern, which supports up to 10 million tokens of context. The release cites only the company's own tests; independent benchmarks have not been published, and outside coverage notes that the result depended on an experienced researcher directing the agents and judging their work.
Why a lab built on research agents draws this much money
Chief executive Behnam Neyshabur worked as a researcher at Google and then Anthropic, as did chief technology officer Harsh Mehta. Their bet is that agents can compress the work of a large research team into a small one, which would lower the cost of training a capable model far below the budgets of the big labs. Mirendil runs on Google Cloud's AI Hypercomputer under a multi-year compute partnership announced in August, using TPUs and NVIDIA systems.
The round arrives as investors fund new labs led by former staff of the largest model makers before they ship much product. Thinking Machines Lab now offers Inkling 2 Small on its Tinker platform, and Reflection AI released open weights for Beam. Anthropic itself is in an IPO roadshow this month. More funding stories are in the AI business hub.
What to watch next
Mirendil has not said when Lor will open beyond the waitlist, what it will cost after the beta, or when larger Mirendil-Base models will finish training. Independent evaluations and API access to its models are promised but have no date. Those results will show whether the 3.5 month, under $10 million claim holds up outside the company.
What changed
| Before | Mirendil funded by a $200 million seed round led by Andreessen Horowitz and Kleiner Perkins, plus an NVIDIA investment |
|---|---|
| After | Mirendil adds a $1 billion Series A led by Kleiner Perkins and Clearlake Capital, with Google among the investors |
| Who it affects | AI research teams on the Lor waitlist; investors in new frontier labs; labs competing for research talent |
Questions people ask
- How much did Mirendil raise?
- $1 billion in a Series A announced on 12 November 2026, led by Kleiner Perkins and Clearlake Capital. The company did not disclose a valuation.
- What is Mirendil building?
- Lor, an agent system in private beta that the company says writes code and runs experiments for days, plus its own models, Mirendil-Base and Fern, aimed at AI research and development work.
- Who founded Mirendil?
- Former Anthropic researchers Behnam Neyshabur, the chief executive, and Harsh Mehta, the chief technology officer.
Sources
- Mirendil Raises $1B to Democratize Frontier AI R&D (Mirendil, 12 Nov 2026)
- Mirendil Raises $1B to Democratize Frontier AI R&D (PR Newswire, 12 Nov 2026)
- Scaling self-accelerating AI with Google (Mirendil, 6 Aug 2026)
- AI Lab Mirendil Raises $1B Series A: Release Names No Valuation, Cites Only Company Tests (Tech Times, 13 Nov 2026)
- Mirendil Raises $1 Billion on a Bet That AI Agents Can Do the Work of an Entire Research Lab (WebProNews, 12 Nov 2026)
Reference pages
Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.
Comments
Loading comments.





