Skip to content

O1-CODER

A coding project that reproduces O1-style reasoning

Visit official site

What O1-CODER is

The project is described as an attempt to reproduce the OpenAI O1 model for programming tasks, using reinforcement learning and Monte Carlo tree search.

What you can do with it

  • Generate standardised test cases with the generator
  • Refine the policy model through self-play and RL
  • Produce code described as more efficient and logical

Getting started

  • Read the background and install guide on GitHub
  • Clone the code base and set up the environment from the README
  • Run the Test Case Generator to produce tests

What to watch out for

  • System-two thinking is the stated target
  • Code quality is assessed through generated tests
  • The methods work in an iterative loop

Pros & cons

✓ What we like

  • Reasoning data comes from self-play
  • Methods work in an iterative refinement loop

! What to watch out for

  • The project is described as an attempt to reproduce O1
  • Code quality is assessed through generated test cases

FAQ

Which techniques are combined?

The material names reinforcement learning and Monte Carlo tree search.

What does the generator produce?

The Test Case Generator makes standardised test cases to assess code.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review