Vector Eval - SWE-Bench: leaderboard
Vector Institute Inspect Evals run of SWE-Bench, measuring agent ability to fix real software issues through code edits and test-based validation.
Metric: Mean Score (%). Source: huggingface.co. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 6.72 | #350 |
| 2 | GPT-4o | 1.2 | #343 |
| 3 | Gemini 1.5 Pro | 0.4 | #445 |
| 4 | O1 | 0.36 | #183 |
| 5 | O3 Mini | 0.24 | #272 |
Interactive version: theaggregate.ai/benchmark?slug=vector-eval-swe-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.