Heads up This site is currently under heavy development.
← all datasets
◆ Evals & benchmarks

Microsoft SWE-bench-Live

MIT

A continuously refreshed SWE-bench built from recent GitHub issues across languages and platforms, so results can't be explained by training-data contamination.

Summary

A brand-new, continuously updated SWE-bench-like dataset powered by an automated curation pipeline. SWE-bench-Live is the first automatically-updating, multi-language and multi-os SWE task set designed for agentic benchmarking and training.
from the project README
Formats
jsonl
License
MIT
Added
2026-09-10

Get it

From the project’s own instructions where it documents any; otherwise a plain clone.

  1. shell
    $ git clone https://github.com/microsoft/SWE-bench-Live.git
  2. shell
    $ cd SWE-bench-Live && pip install -e .

    Python 3.10 or later. The task instances are on Hugging Face; evaluation runs each one in Docker.

Contents

37 files 2.2 MB repository

  • .py 21
  • .md 8
  • (no ext) 3
  • .png 2
  • .sh 2
  • .toml 1

Use it with

Commands are curated, not yet run by us.

jq — Stream the records one per line to see their shape.

sample.jsonl
$ jq -c . sample.jsonl

DuckDB — Query the file in place with SQL — DuckDB detects CSV, JSON and JSONL by extension.

sample.jsonl
$ duckdb -c 'SELECT * FROM '\''sample.jsonl'\'' LIMIT 10'
my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →

    Send us feedback