whatisgithub

What is s1?

ludvik/s1 — explained in plain English

Analysis updated 2026-07-18 · repo last pushed 2025-02-18

Audience · researcherComplexity · 5/5StaleSetup · hard

In one sentence

Trains an AI model to reason step-by-step through hard problems using just 1,000 examples and a technique called budget forcing.

Mindmap

mindmap
  root((repo))
    What it does
      Step-by-step reasoning
      Budget forcing
      1000 training examples
    Tech stack
      32B model
      Hugging Face
      Training scripts
    Use cases
      Math problem solving
      Coding challenges
      Reproducible reasoning
    Audience
      Researchers
      Developers
    Extras
      Published paper
      s1.1 improved version
      Eval scripts included

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Train your own step-by-step reasoning model instead of relying on a proprietary one.

USE CASE 2

Download the pretrained 32B model from Hugging Face and run inference with a few lines of code.

USE CASE 3

Use budget forcing to control how long the model thinks before answering.

USE CASE 4

Evaluate the model on math or coding problems using the included evaluation scripts.

What is it built with?

PythonHugging FacePyTorch

How does it compare?

ludvik/s10verflowme/alarm-clock0verflowme/seclists
Language—CSS—
Last pushed2025-02-182022-10-032020-05-03
MaintenanceStaleDormantDormant
Setup difficultyhardeasyeasy
Complexity5/52/51/5
Audienceresearchervibe coderops devops

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · hard Time to first run · 1day+

Training recommends 16 high-end GPUs, inference is easier via Hugging Face.

Not stated in the explanation.

Copy-paste prompts

Prompt 1
Show me how to download the s1 model from Hugging Face and run a basic inference example.
Prompt 2
Explain how budget forcing controls the model's reasoning length in this project.
Prompt 3
Walk me through the training code to understand how the 1,000 example dataset is used.
Prompt 4
Compare the original s1 model with the improved s1.1 version using r1 reasoning traces.

Frequently asked questions

What is s1?

Trains an AI model to reason step-by-step through hard problems using just 1,000 examples and a technique called budget forcing.

Is s1 actively maintained?

Stale — no commits in 1-2 years (last push 2025-02-18).

What license does s1 use?

Not stated in the explanation.

How hard is s1 to set up?

Setup difficulty is rated hard, with roughly 1day+ to a first successful run.

Who is s1 for?

Mainly researcher.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.

What are similar repos?