Anthropic Alumni Startup Mirendil Raises $1 Billion Series A, Betting on Agent-Driven Model Development

Mirendil, a San Francisco startup founded by two former Anthropic researchers, announced a $1 billion Series A round on November 12, led by Kleiner Perkins and Clearlake Capital. The round comes 141 days after a $200 million seed announced in June, bringing total funding to $1.2 billion in under five months, according to FourWeekMBA's analysis of the company's PR Newswire release.

The release names Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 as additional participants. It does not disclose a valuation, the split between new and existing investors, or how much each lead contributed. A database tracking AI funding rounds lists the seed at a $1 billion valuation and the Series A as reported at a $5 billion valuation, though the latter remains unconfirmed by the company or a regulator. The same database records the Series A date as September 22 and lists Andreessen Horowitz as a lead investor, conflicting with the November 12 announcement date in the company release and the release's omission of Andreessen Horowitz from the investor list.

What Mirendil Claims to Have Built

Mirendil says it is "building tools to democratize frontier AI R&D and help put it within reach of every scientist and engineer." The release describes three products:

• Mirendil-Base: A coding model the company says was pretrained from scratch by "one researcher, working with agents" in 3.5 months using "less than $10M in compute."
• Fern: A post-trained version of Mirendil-Base that "holds up to 10M tokens of context."
• Lor: An agent product described as now in private beta that runs hundreds of agents writing code and running experiments for days at a time, with the company stating it runs on Lor daily.

No independent benchmarks for Mirendil-Base or Fern are provided. Lor has no stated launch date, pricing, or customer names. The release gives no headcount, revenue, or traction metrics, describing the company only as "still a small team with a singular focus."

Founders' Background

Co-founders Behnam Neyshabur (CEO) and Harsh Mehta met at Google in 2019, moved to Anthropic in late 2024, and left in December 2025, according to a June report from The Next Web cited by FourWeekMBA. Neyshabur spent over five years at Alphabet co-leading reasoning research for Gemini; Mehta built the first version of Anthropic's internal AI-research platform, per the same report. At the seed stage, Mirendil had roughly 20 researchers and engineers drawn from Anthropic, xAI, Google DeepMind and OpenAI, The Next Web reported.

Investment Continuity and Context

Kleiner Perkins is the only firm named as a lead in both the seed and Series A rounds. The seed was co-led by Andreessen Horowitz and Kleiner Perkins, with Nvidia participating; the Series A release does not name Andreessen Horowitz or Nvidia, though a funding database lists Andreessen Horowitz as a Series A lead with "Reported" status. Whether either firm participated in the new round is not established by the release.

The $1 billion Series A is at least 100 times the under-$10 million of compute Mirendil says it spent pretraining its first model, by FourWeekMBA's arithmetic. From the founders' December 2025 departure from Anthropic to this round is about 11 months.

What Is Not Established

The release does not state a valuation, the amount each lead invested, the new-versus-existing investor split, or a launch timeline or price for Lor. It does not confirm whether Andreessen Horowitz or Nvidia rejoined. The founders' backgrounds come from a June media report, not the release itself. Mirendil was not contacted for the FourWeekMBA analysis.

The company's central claim — that agents can shrink the team needed for frontier model development by orders of magnitude — rests on its own, unverified compute figure for a single model with no published benchmarks. Lor remains in private beta with no external validation. The fundraising pace and scale reflect investor conviction in the founders' pedigree and the agent harness thesis, not demonstrated product traction or revenue.

SOURCES:

[2] fourweekmba.com: Mirendil Raises $1B Series A, 5x Its Seed in Under 5 Months URL: https://fourweekmba.com/ai-mirendil-raises-1b-series-a-5x-its-seed-in-under-5-months

[3] everyneo.com: AI lab funding rounds: 179 rounds with status and sources URL: https://everyneo.com/rounds