Breaking
$2B RAISED at an $8B valuation - Series B, Oct 2025 ASIMOV reads your Slack, not just your code 15x valuation jump in seven months NVIDIA + SEQUOIA + LIGHTSPEED on the cap table 60 people. One very large ambition. OPEN WEIGHTS - the Western answer to DeepSeek $2B RAISED at an $8B valuation - Series B, Oct 2025 ASIMOV reads your Slack, not just your code 15x valuation jump in seven months NVIDIA + SEQUOIA + LIGHTSPEED on the cap table 60 people. One very large ambition. OPEN WEIGHTS - the Western answer to DeepSeek
Company Profile / Frontier AI

Reflection AI

A 60-person lab in lower Manhattan is trying to open-source superintelligence - and starting with the thing engineers actually do all day: write code.

Reflection AI logo and wordmark
The wordmark: one hollow circle, one filled. Somewhere between a logic gate and a pair of eyes. Photographed off the company's own front page.
$8B
Valuation
$2.13B
Total raised
2024
Founded
~60
Employees
Share this LinkedIn Twitter / X Facebook Instagram

The lab that raised two billion dollars before shipping a model

Walk into the conversation about American AI in late 2025 and Reflection AI is the name that does not fit the pattern. It has no consumer chatbot. It has not released the frontier model it keeps promising. And yet, in October 2025, investors handed it $2 billion at an $8 billion valuation - one of the largest early-stage bets the industry has ever placed.

What they bought was not a product. It was a thesis, two founders, and what the company calls the highest-density reinforcement-learning talent of any startup. The thesis is blunt: whoever teaches machines to write software well enough will have built the engine for everything else. Reflection wants to be that lab, and it wants to do it in the open.

"We built something once thought possible only inside the world's top labs."

- Misha Laskin, Co-Founder & CEO

Frontier AI had quietly become a private club

By 2024 the most capable models lived behind locked doors. OpenAI, Anthropic and Google held the weights, the data and the keys. If you were a researcher, a hospital, or a government that wanted to build on a frontier model, your options were to rent access on someone else's terms - or to download an open model from China.

That second option is the part that kept people up at night. DeepSeek had shown the world that excellent open-weight models could come out of Beijing. The uncomfortable question: why was the leading open frontier model not American? Reflection's answer was to simply go build one.

"Open weights, closed datasets. They want to be the answer to DeepSeek - made in New York."

- The thesis, in one line
Reporter's note: "Open" here has fine print. Reflection plans to release model weights publicly while keeping its datasets and training pipelines proprietary. Free for researchers; paid for enterprises and governments. Open enough to matter, closed enough to bill.

Two people who had already built the impossible

Misha Laskin led reward modeling for DeepMind's Gemini and trained under Pieter Abbeel at Berkeley. Ioannis Antonoglou spent more than a decade at DeepMind and was a core architect of AlphaGo - the system that beat a human world champion at Go in 2016 and convinced a generation that reinforcement learning was not a toy.

They left to make a specific wager: that coding is the cleanest path to general superintelligence. Code can be run. It can be tested. Success and failure are not matters of opinion - they compile or they do not. That feedback loop is exactly what reinforcement learning feeds on. Solve coding, the bet goes, and you have built a machine that can improve itself.

"They believe solving autonomous coding is the on-ramp to superintelligence. The compiler is the referee."

- On the founding wager

Asimov, the agent that reads the room

In July 2025 the company shipped its first public product: Asimov, named for the writer who gave robots three laws. Most coding tools autocomplete your next line. Asimov tries to understand why the software exists at all.

It ingests the code, yes - but also the documentation, the project notes, the emails, the Slack threads where the real decisions were actually made. Under the hood, a swarm of small long-context "retriever" agents pull relevant fragments from a sprawling codebase, and a single large "combiner" agent reasons over them to produce one coherent answer. A feature called Asimov Memories lets a team's most senior engineers offload what lives in their heads so the whole team inherits it.

Asimov

Autonomous code comprehension agent for large codebases. Multi-agent retriever/combiner architecture. Learns from code, docs, Slack and email - not just commits.

Open frontier model

A Mixture-of-Experts language model trained on tens of trillions of tokens, weights to be released publicly. Targeted for early 2026; text first, multimodal later.

RL post-training stack

Reinforcement-learning techniques that let models plan, debug, test and refactor whole systems - and learn from their own failures - rather than predict one line at a time.

Caption, for the skeptics: In early blind tests, codebase maintainers preferred Asimov's answers between 60% and 80% of the time. Which means roughly a third of the time, the humans still wanted the human. Progress, not magic.

Eighteen months, fifteen-fold

2024
Founded in New York by Misha Laskin and Ioannis Antonoglou, fresh out of Google DeepMind. A $25M seed from Sequoia and CRV gets it going.
MAR 2025
Public launch with $130M in early funding (Lightspeed, Sequoia, CRV, with NVIDIA). Valuation around $545M.
JUL 2025
Asimov ships - a code-comprehension agent that learns from far more than source code.
OCT 2025
The $2B round. Series B at an $8B valuation - a roughly 15x jump in seven months. NVIDIA, Sequoia, Lightspeed, Citi, Eric Schmidt and others pile in.
EARLY 2026
Frontier model targeted for release - open weights, trained on tens of trillions of tokens.

A valuation that refused to sit still

Reflection AI implied valuation, by round
Mar 2025
Series A
$0.55B
Oct 2025
Series B
$8B
Capital raised, cumulative
Through
launch
$130M
After
Series B
$2.13B total

Bars scaled to the largest figure in each group. Figures from reported funding announcements, Oct 2025. The Series A valuation is widely reported as roughly $545M.

Why build it in the open at all

The easy money in AI is in keeping the secrets. Reflection is wagering the opposite - that the lab which hands out the most capable open weights ends up at the center of the ecosystem, the way the most useful public square ends up owning the conversation. Researchers build on you for free. Enterprises and governments pay to deploy you privately and to run sovereign systems they control.

It is a bet on gravity rather than secrecy. Whether it pays is the open question hanging over the whole company.

// founders

AlphaGo's co-architect and Gemini's reward-modeling lead, in one room.

// the name

Asimov, after the author of the Three Laws of Robotics. Subtle.

// the swarm

Many small retriever agents feed one big reasoner. Divide, then conquer.

// the backers

NVIDIA, Sequoia, Lightspeed, Citi, GIC, Eric Schmidt, Eric Yuan.

// the scale

Tens of trillions of tokens going into a Mixture-of-Experts model.

// the rivals

OpenAI and Anthropic on one side, DeepSeek and Llama on the other.

Back to the lab that raised two billion dollars

Return to where this started: a company valued at eight billion before its headline product exists. From the outside it looks like a bet on a slide deck. Up close it is a bet on two people who have already done the once-impossible twice, and on a wager that the open path to superintelligence runs straight through the boring, testable, endlessly verifiable work of writing software.

If Reflection is right, the frontier model it ships in 2026 becomes a public utility that anyone - a student, a startup, a state - can build on without asking permission. If it is wrong, it becomes the most expensive thesis in recent AI history. Either way, the experiment is now funded, staffed, and running. The compiler will be the judge.

"Most labs sell you the answer. Reflection wants to hand you the engine and let you ask your own questions."

- The closing argument