Mirendil Raises $1 Billion, Without Disclosing Its Valuation

By Olivia Hart

Mirendil announced a $1 billion Series A on November 12, according to FourWeekMBA's analysis of the company's PR Newswire release. FourWeekMBA places the Series A 141 days after the June announcement of a $200 million seed. Together, the rounds brought total funding to $1.2 billion in under five months.

Mirendil's release names Kleiner Perkins and Clearlake Capital as Series A leads for the San Francisco startup. It also names Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 as participants. The release discloses no valuation, no contribution amount from either lead and no split between new and existing investors.

What does Mirendil say it has 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 but provides no independent benchmarks for Mirendil-Base or Fern.

• Mirendil-Base: A coding model that the company says was pretrained from scratch by "one researcher, working with agents" in 3.5 months, using "less than $10M in compute".
• Fern: This post-trained version of Mirendil-Base "holds up to 10M tokens of context", according to Mirendil. The release cites no independent benchmark for that capacity claim.
• Lor: This agent product, now in private beta, runs hundreds of agents writing code and running experiments for days at a time. Mirendil says it runs Lor daily, but the release gives no launch date, pricing or customer names.

The release provides no headcount, revenue or traction metrics. Mirendil describes itself only as "still a small team with a singular focus", leaving the scale of its commercial operation unmeasured.

The immediate takeaway is plain: Mirendil has described products, not demonstrated their performance outside the company.

Who is behind the startup?

Co-founders Behnam Neyshabur, the CEO, and Harsh Mehta are former Anthropic researchers who met at Google in 2019, according to a June report from The Next Web cited by FourWeekMBA. The report says both moved to Anthropic in late 2024 and left in December 2025.

Neyshabur spent more than five years at Alphabet, co-leading reasoning research for Gemini. Mehta built the first version of Anthropic's internal AI-research platform, The Next Web reported.

At the seed stage, Mirendil had roughly 20 researchers and engineers drawn from Anthropic, xAI, Google DeepMind and OpenAI, The Next Web reported. The founders' backgrounds come from that June media report rather than the fundraising release.

FourWeekMBA says the fundraising pace and scale reflect investor conviction in the founders' pedigree and the agent harness thesis, not demonstrated product traction or revenue. The plain takeaway: pedigree attracted the capital, but customer demand remains undocumented.

Which investors backed the rounds?

Kleiner Perkins is the only firm the company release names as a lead in both rounds. The June seed was co-led by Andreessen Horowitz and Kleiner Perkins, with Nvidia participating, according to FourWeekMBA's analysis. The Series A release does not name Andreessen Horowitz or Nvidia, so their participation remains unestablished.

EveryNeo's AI funding-round database lists a reported $5 billion Series A valuation, unconfirmed by Mirendil or a regulator. It also records the seed at a $1 billion valuation and dates the Series A to September 22, conflicting with the company's November 12 announcement. The database lists Andreessen Horowitz as a lead investor with "Reported" status.

The plain takeaway is continuity at Kleiner Perkins, but incomplete disclosure about the rest of the investor group.

What remains unproven?

Mirendil's central claim is that agents can shrink the team needed for frontier model development by orders of magnitude. FourWeekMBA says that claim rests on the company's unverified compute figure for one model, with no published benchmarks. Lor remains in private beta without external validation.

FourWeekMBA calculates the $1 billion Series A at least 100 times the under-$10 million compute Mirendil says it spent pretraining its first model. The financing came about 11 months after the founders' December 2025 departure from Anthropic.

FourWeekMBA did not contact Mirendil for the analysis. The release also omits valuation, individual lead contributions, the new-versus-existing investor split, and any launch date or price for Lor.

The next proof points are specific: independent model benchmarks, Lor's pricing and launch schedule, and disclosed revenue or customer traction. Until they arrive, the $1.2 billion in funding proves investor appetite, not commercial execution.