Mirendil Raises $1B Series A, Reportedly at a $5B Valuation
Kleiner Perkins and Clearlake Capital led the Series A. Lor, Mirendil's agent, is free in beta, and an investor's hiring post says Mirendil wants a first go-to-market lead to turn pilots into paid deals.
TL;DR
- Mirendil raised a $1 billion Series A led by Kleiner Perkins and Clearlake Capital, reportedly at about a $5 billion valuation.
- The round came 141 days after a $200 million seed at $1 billion. Lor, its agent product, is still free and invite-only.
- The next markers are Mirendil's founding go-to-market hire and whether Lor carries a price after the beta.
Mirendil, which builds AI agents that run research work for AI teams, has raised $1 billion.
Kleiner Perkins and Clearlake Capital led the Series A, Mirendil said on November 12. Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 joined. The company's post names no valuation. An Andreessen Horowitz hiring newsletter put it at $5 billion. The firm, known as a16z, backed Mirendil's seed 141 days earlier, a $200 million round at $1 billion.
Lor, Mirendil's agent product, is free and invite-only during its beta. The same newsletter says Mirendil is hiring a founding go-to-market lead to "find and win Mirendil's customers." That person would run hands-on pilots with researchers and engineers and turn them into paid deals. Mirendil also wants a founding finance hire as it starts billing its first enterprise customers.
Lor runs agents that work for days at a time
Lor's agents write code, run experiments, provision their own compute and keep working for days, according to Mirendil's launch post. They run on models from Anthropic, OpenAI, Google, xAI and open-source developers, alongside Mirendil's own. "We're building Lor to do the work of an entire AI lab," the company wrote.
In one of Mirendil's own runs, 102 agents built environments from 100 PyTorch pull requests. In another, agents working overnight found and fixed security bugs in LiteLLM.
Mirendil says one researcher working with agents pretrained its Mirendil-Base model in 3.5 months, for under $10 million in compute. With agents doing the hands-on work, "the size of your team stops limiting what you can build," the company wrote. The a16z newsletter puts the team at about 20 researchers and engineers.
Mirendil's own tests carry the evidence so far
Tech Times points out that the release cites no outside evaluation of any Mirendil model. The under-$10 million bill covers only Mirendil-Base, the smallest model. Larger models, from 0.5 trillion to 10 trillion parameters, were still training on October 21.
Google is both an investor in the Series A and Mirendil's compute supplier, under a partnership signed in August. The a16z newsletter values that partnership at more than $100 million.
Setting $1 billion against a sub-$10 million training bill measures the round against the smallest model in the family. Tech Times compares it with other AI labs instead. Thinking Machines Lab raised about $2 billion at $12 billion in July 2025, and Recursive Superintelligence raised $650 million at $4.65 billion in May.
Mirendil now has to sell a beta it gives away free
Mirendil's pitch is that a small team with agents can do a lab's work. Lor has no published price, and Mirendil is still hiring the go-to-market lead who would turn pilots into paid deals. Its larger models were still training on October 21.
We're building Lor to do the work of an entire AI lab
Mirendil's larger models, from 0.5 trillion to 10 trillion parameters, were still training on October 21, and Lor remains free and invite-only in its beta.
