benchmarks.wiki / Public workspace

SWE-bench Verified

Code / Agents

SWE-bench Verified is the 500-issue subset of SWE-bench that human annotators confirmed to be solvable and correctly specified. Each task is a real GitHub issue in a Python repository.

swe-bench-verified

Official source

History

SWE-bench Verified was released in August 2024 through work by OpenAI and the SWE-bench team to improve the reliability of repository-level software engineering evaluation. It selected tasks from the original SWE-bench collection after human review of whether their issue descriptions and tests supported a fair, solvable task. The dataset is hosted under princeton-nlp on Hugging Face.

How it was built

The benchmark contains 500 tasks drawn from real GitHub issues in Python repositories. Each task is tied to a base commit and a test-based evaluation of a candidate patch. Human annotators screened the tasks for solvability and specification quality, reducing cases where an unclear request or unsuitable tests could confound the result. Successful patches must fix the targeted failing tests while preserving the required previously passing tests.

At a glance

What a problem looks like

A real GitHub issue with the repository, base commit and the tests that must go from failing to passing.

How it is scored

The dataset declares no licence, and a run needs the repository checkout and test harness rather than the issue text alone. The catalogue links to the official distribution.

Reported scores

Evaluation conditions differ between rows, so this is not a ranking. Rows marked vendor are reports by the model’s own organization.

ModelScoreMetricSetupReported bySource
Claude 4.5 Opus (20251101) (medium)Anthropic74.4resolved (%)Verified; mini-SWE-agent; Model: claude-opus-4-5-20251101; Org: Anthropic; System: Attempts - 1; Mini: 1.16.0leaderboardVerified leaderboard; 20251124_mini-v1.16.0_claude-opus-4-5-20251101Reported 2025-11-24
Gemini 3 Pro Preview (2025-11-18)Google DeepMind74.2resolved (%)Verified; mini-SWE-agent; Model: gemini-3-pro-preview; Org: Google DeepMind; System: Attempts - 1; Mini: 1.15.0leaderboardVerified leaderboard; 20251118_mini-v1.15.0_gemini-3-pro-preview-20251118Reported 2025-11-18
OpenHands + GPT-5OpenAI71.8resolved (%)Verified; OpenHands; Model: openai/gpt-5-2025-08-07; Org: OpenHands; System: Attempts - 1leaderboardVerified leaderboard; 20250807_openhands_gpt5Reported 2025-08-07

Why it is hard

A patch must make specific tests pass without breaking others, inside a real codebase the model has not seen.

Limitations and criticism

A resolved task means that the patch satisfies the evaluation tests, not that it meets every unstated user expectation or is ready for production. The collection covers a particular set of public Python repositories, so it does not represent all languages or software engineering work. Public issues, historical patches and test data create potential prior-exposure risks. Agent tools, retry budgets and scaffolding also affect results, making the reported score a property of the complete evaluation system.

How much room is left

The leaderboard records included here report 74.4 percent resolved for Claude 4.5 Opus at medium effort and 74.2 percent for Gemini 3 Pro Preview, using mini-SWE-agent in November 2025. These results leave roughly a quarter of the tasks unresolved in the reported configurations. The records use different mini-SWE-agent versions, so even their close scores should not be read as a controlled comparison of models alone. They demonstrate remaining headroom in those reports rather than establishing the current leaderboard ceiling.

Discussion

1 thread and 0 posts on this page. Anyone may reply; no account is required.

Contribute

Share a reproduction, report an evaluation issue, or suggest a correction to this benchmark overview.

Public problem summary

Catalogue metadata only; public problem statistics are not available.

Public problems
—
Formats
—
Problems with images
—
Answer availability
—
Configs
—
Splits
—
Awaiting a first discussion
—
Problems with a discussion
—
Public contributions
—