> ## Documentation Index
> Fetch the complete documentation index at: https://rllm-org-rllm-19-feat-renderer-parser-backend.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# DeepSWE

> A 32B software engineering agent achieving 59% on SWE-Bench-Verified

<Card title="Read the full write-up" icon="arrow-up-right-from-square" href="https://pretty-radio-b75.notion.site/DeepSWE-Training-a-Fully-Open-sourced-State-of-the-Art-by-Scaling-RL-22281902c1468193aabbe9a8c59bbe33">
  Notion blog post with full details
</Card>

DeepSWE is a 32B software engineering agent that achieves **59% on SWE-Bench-Verified** with test-time scaling (42.2% Pass\@1). It tops the SWE-Bench leaderboard for open-weight models.

## Results

| Model       | Parameters | SWE-Bench-Verified |
| ----------- | ---------- | ------------------ |
| **DeepSWE** | **32B**    | **59.0%**          |

## Approach

Trained on top of Qwen3-32B to search, view, and navigate codebases, using RL to improve software engineering task completion.

See [`examples/harbor_swe`](https://github.com/rllm-org/rllm/tree/main/examples/harbor_swe) for the unified-trainer SWE-bench reproducer.

**Released**: July 2025