SWE-Cycle: leaderboard

SWE-Cycle (2026) tests code agents on the full issue-resolution cycle: from a raw repository and an issue they must build the environment, implement the fix and write tests in one session, over 489 filtered SWE-bench issues. The score is the percent of static-review and execution checks passed across the three phases as graded by an agentic judge, higher is better.

Metric: FullCycle score (%): the mean of the static and dynamic scores (SWE-Judge (Claude Opus 4.5 agent) static code review and dynamic execution, each scored 0, 1 or 2, normalized to 0-1) of the environment, implementation and test phases, on the 489 SWE-Cycle issues (225 SWE-bench Verified, 203 SWE-bench Pro, 61 SWE-bench Multilingual, filtered for contamination, lifecycle complexity and reliable tests); FullCycle task: from a raw codebase and the issue, set up the environment, implement the fix and write tests in one session (3 hours); agents run in the OpenCode harness in isolated Docker containers; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GLM-5.181.49#198
2Claude Sonnet 4.680.52#173
3GPT-5.479.94#135
4SWE-Cycle Qwen-3.5 (checkpoint unspecified)74.44
5Kimi K2.570.51#266
6MiniMax-M2.739.84#539

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=swe-cycle · How It Works · Data refreshed daily, snapshot 2026-11-21.