Dashboard

Get started on Prime Intellect

Start with the Prime CLI and your coding agent. Add billing later, once you are ready to launch hosted jobs or rent compute.

1

Install and Sign In

Use browser login to connect the CLI to your Prime account. No billing setup is required for this step.

Install Prime CLI

uv tool install -U prime

Sign in

prime login
2

Prepare Your Workspace

Run setup once from your project root. It prepares your selected coding agent, starter configs, and Prime endpoints.

Set up workspace

prime lab setup

During setup

Once you run setup, you will be asked which coding agent you primarily use. This will not open your agent; it just ensures the best instructions are set up locally. Press Enter to use Codex, or type your primary agent.

You will then be prompted to set up skills for any other agents you use.

3

Run a Quick Evaluation

Trigger a one-example eval from the CLI so you can confirm your account, endpoint, environment, and result upload path all work.

Quick test

Small text task that should run end to end with Prime Inference.

prime eval run primeintellect/reverse-text -m openai/gpt-oss-20b -p prime -n 1 -r 1 -t 512 -s -A

Reasoning example

Single AIME problem with a higher token cap so the model can finish its answer.

prime eval run primeintellect/aime2026 -m openai/gpt-oss-20b -p prime -n 1 -r 1 -t 2048 -s -A

After the reasoning example, choose another environment from the Environments Hub.

4

Hand a Task to Your Coding Agent

Open your agent of choice and give it a starting task. It'll know how to use the Prime stack.

Pick a starting task, then paste the generated prompt:

I want to train a model for math reasoning. Start from the closest Prime starter recipe in configs/rl, run a baseline eval, then modify the existing environment or reward for this task. Keep the first change small, run another eval after training, and summarize what improved or regressed.
5

Scale Up When You Need It

Add billing only when you are ready for hosted workloads or rented compute.

Evaluations

Benchmark models locally or on hosted infrastructure.

Training

Launch managed training runs with logs and run tracking.

On-demand GPUs

Deploy compute when local iteration is not enough.

Environments Hub

Discover, upload, and run Prime environments.

Rate your experience