Give Primus a research problem and it delivers results like a real ML team, whether that's a ported model, an optimized kernel, a forecaster or an eval suite. Every result ships with a full research paper and a reproducibility package including code, artifacts and logs.

Read Primus' research →

Getting started​

During this launch period, GPU usage is on us.

1. Sign up at primus.lab.cloud.
2. Add credits to run experiments. Credits cover model tokens. GPU usage is free for now.
3. Ask Primus the questions your ML team has been most wanting to answer. Writing research prompts has tips for your first one.

Your account starts on the Basic plan. If you want access to better GPUs and models, or want to run multiple projects in parallel, you can upgrade to the Plus plan.

The docs walk through your first project, and the FAQ covers common questions. If you're an organization that wants to bring Primus to your team, get in touch.

Get started with Primus →

Every innovation we rely on today comes from research. Experts read the literature, propose a theory, run the experiments and publish their findings. This research loop, until now, has run at human speed where one turn could take months.

Primus runs the same loop 30× faster. It autonomously hypothesizes, reads millions of papers, codes, runs experiments on real compute, learns and iterates, draws conclusions, delivers artifacts and writes the final paper. The more experiments Primus runs, the better it gets.

We ran Primus in our lab for 30 days. It wrote 30 complete papers, each on a question never before answered across fields like LLMs, seismology, physics, crystallography and biology.

Findings that Primus helped produce have already been cited in published research from labs like DeepMind.

Read Primus' research →

What researchers are saying​

Here's feedback from researchers that have used Primus and reviewed its output:

"These papers reflect what I'd expect from a strong master's-level researcher with two to three years of engineering experience. I estimate it would take a full-time researcher roughly eight to twelve weeks per paper, covering everything from problem identification and research ideation to experimental design and preparing the final manuscript."

— Shashank Shekhar

"We had early access to Primus at Rubric, and I can confidently say we have never experienced anything like it. We're a small team, and we have plenty of ideas that get shelved because we lack the resources to execute them. With Primus, it feels like adding a team of world-class, senior ML engineers to your roster overnight who can chip away at your ideas. The quality of the research and output is in a league of its own. It truly feels like execution at the speed of thought."

— Sarim Malik, CEO, Rubric Labs

"Primus collapses months of human experimentation, literature review, and drafting into a single, automated workflow. The implications for R&D velocity and cost-efficiency are massive."

— Jash Mehta, Applied Research Scientist, Hippocratic AI

What's next​

Primus is the first step towards our goal to accelerate innovation in every experimental domain and to put that power in the hands of anyone who does research, not just a few frontier labs.

Primus is officially live. To ensure a seamless experience despite limited GPU capacity, we are gradually onboarding users from our waitlist starting today. Join the waitlist to secure your spot and we will grant you access as quickly as we can. If you're an organization that wants to bring Primus to your team, get in touch.

Let's discover the unknown together.

Get started with Primus →