ChainSWE (Sequential): leaderboard
Metric: Per-bug resolution rate (%; share of the 304 bugs in 100 chronological bug chains from six SWE-bench-family datasets whose tests pass, each bug starts from the repository the agent left after the earlier bugs, with a fresh conversation; the Baseline configuration of the SWE-Edit scaffold, which the paper states is identical to the original SWE-agent; 100 turns and 30 minutes per bug, thinking effort medium). Source: arxiv.org. Saturation forecast: Around 2033. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 49 |
| 2 | GPT-5.4 Nano (Medium) | 45.1 |
| 3 | GPT-5.4 Mini (Medium) | 41.8 |
| 4 | Claude Opus 4.7 (Medium) | 40.5 |
| 5 | Gemini 3.1 Pro (Preview) (Medium) | 36.5 |
Interactive version: theaggregate.ai/benchmark?slug=chainswe-sequential · How It Works · Data refreshed daily, snapshot 2026-09-29.