SWE-bench

📣 New benchmark: CodeClash (website, github) evaluates SWE agents on goals, not tasks
📣 New: Meet mini, the 100 line AI agent that still gets 65% on SWE-bench verified!

Software engineering agents, benchmarks, and models.

Built and maintained by researchers from Stanford University and Princeton University.

This organization contains the source code for several projects in the SWE-* open source ecosystem, including:

SWE-bench, a benchmark for evaluating AI systems on real world GitHub issues.
SWE-agent, a system that automatically solves GitHub issues using an LM agent.
SWE-smith, a toolkit for generating SWE training data at scale.
mini, an AI agent written in just 100 lines of code that scores >70% on SWE-bench verified

Also check out the supporting infrastructure for working with SWE-* projects

SWE-ReX, infrastructure supporting sandboxed code execution for AI agents
sb-cli, a command line interface for running evaluations on the cloud.
Mirror clones for the SWE-bench and SWE-smith repositories are available here and here.

Pinned Loading

SWE-bench: Can Language Models Resolve Real-world Github Issues?

Python 5.2k 894
[NeurIPS 2025 D&B Spotlight] Scaling Data for SWE-agents

Python 678 120
Open sourced predictions, execution logs, trajectories, and results from model inference + evaluation runs on the SWE-bench task.

Shell 269 312
Run SWE-bench evaluations remotely

Python 70 9

Repositories

Showing 9 of 9 repositories

SWE-smith Public
[NeurIPS 2025 D&B Spotlight] Scaling Data for SWE-agents

SWE-bench/SWE-smith’s past year of commit activity
SWE-bench Public
SWE-bench: Can Language Models Resolve Real-world Github Issues?

SWE-bench/SWE-bench’s past year of commit activity
experiments Public
Open sourced predictions, execution logs, trajectories, and results from model inference + evaluation runs on the SWE-bench task.

SWE-bench/experiments’s past year of commit activity
SWE-bench/swe-bench.github.io’s past year of commit activity

JavaScript
13 15 5 4
Updated Mar 29, 2026
SWE-smith-envs Public
Artifacts for building environments (Docker images) for repositories represented in SWE-smith

SWE-bench/SWE-smith-envs’s past year of commit activity

Dockerfile
5 2 0 0
Updated Mar 2, 2026
reading-list Public
Academic papers and works related to SWE-bench and SWE-agents

SWE-bench/reading-list’s past year of commit activity

11 4 0 0
Updated Dec 8, 2025
SWE-bench/.github’s past year of commit activity

0 MIT 0
0 0
Updated Nov 14, 2025
sb-cli Public
Run SWE-bench evaluations remotely

SWE-bench/sb-cli’s past year of commit activity

Python
70
MIT
9 13 1
Updated Aug 14, 2025
humanevalfix-results Public archive
Evaluation data + results for SWE-agent inference on HumanEvalFix task

SWE-bench/humanevalfix-results’s past year of commit activity

Jupyter Notebook
1
0
0 0
Updated Jul 11, 2024