Skip to main content
Prime Intellect

​
Welcome to Verifiers

Verifiers is a Python library for creating environments to train and evaluate LLMs with reinforcement learning. Build custom RL environments with datasets, reward functions, and interaction harnesses for training capable language model agents.

Quick Start

Get up and running in minutes with your first environment

Installation

Install Verifiers with pip, uv, or poetry

GitHub

View the source code and contribute

Environments Hub

Browse and share environments with the community

​
What are Environments?

Environments contain everything required to run and evaluate a model on a particular task:
  • Dataset - Task inputs for the model to solve
  • Harness - Tools, sandboxes, context management, and interaction protocols
  • Rubric - Reward functions to score the model’s performance
Environments can be used for:
  • Training models with reinforcement learning (RL)
  • Evaluating model capabilities
  • Generating synthetic data
  • Experimenting with agent harnesses
  • Creating custom benchmarks

​
Key Features

Multiple Environment Types

From simple single-turn Q&A to complex multi-turn agents with tool use

Flexible Reward Functions

Define custom rubrics with multiple reward functions and metrics

Tool Integration

Stateless tools, MCP servers, sandboxed execution, and browser automation

Seamless Training

Integrated with prime-rl for distributed RL training

​
Simple Example

Create a basic math environment in just a few lines: 
import verifiers as vf
from datasets import Dataset

def load_environment():
    # Your task data
    dataset = Dataset.from_list([
        {"prompt": [{"role": "user", "content": "What is 2+2?"}], "answer": "4"},
        {"prompt": [{"role": "user", "content": "What is 3*5?"}], "answer": "15"},
    ])
    
    # Your reward function
    async def correct_answer(completion, answer) -> float:
        response = completion[-1]["content"]
        return 1.0 if answer in response else 0.0
    
    rubric = vf.Rubric(funcs=[correct_answer])
    
    return vf.SingleTurnEnv(dataset=dataset, rubric=rubric)

​
Ecosystem Integration

Verifiers is tightly integrated with the Prime Intellect platform:

​
Get Started

Installation Guide

Install Verifiers and the Prime CLI

Quick Start Tutorial

Build and run your first environment

​
Latest Updates

v0.1.9 (January 2026) - New experimental environment classes, monitor rubrics for automatic metric collection, improved workspace setup, better error handling, and documentation overhaul.
  • v0.1.9 - Experimental environment classes, monitor rubrics, improved workspace setup
  • v0.1.8 - Trajectory-based rollout tracking for token-in token-out training
  • v0.1.7 - Improved quickstart, new RLTrainer, documentation improvements

​
Citation

If you use Verifiers in your research, please cite: 
@misc{brown_verifiers_2025,
  author       = {William Brown},
  title        = {{Verifiers}: Environments for LLM Reinforcement Learning},
  howpublished = {\url{https://github.com/PrimeIntellect-ai/verifiers}},
  year         = {2025}
}
Originally created by Will Brown (@willccbb).

Build docs developers (and LLMs) love

Get started for freeTalk to us