TEBench - Stale Test Update: leaderboard
Update score (%) on tasks with stale tests on TEBench (2026), a project-level test-evolution benchmark of 314 commits from 10 Java projects in which an agent in its default configuration updates the test suite after a production-code change. The score gates on executability and blends coverage overlap and modification similarity; higher is better.
Metric: Update overall composite (%) for tests left stale by the change: executability gate times a blend of coverage overlap with the developer tests on the changed methods and token-level modification similarity, on TEBench (314 commits from 10 Java projects): given the project at the commit with its production changes but not its test changes, the agent edits the test suite in an isolated worktree with default agent settings; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.3 Codex | 70.8 | #124 |
| 2 | GLM-5 | 66.7 | #269 |
| 3 | Claude Sonnet 4.6 | 65.4 | #173 |
| 4 | Qwen3.5 (TEBench checkpoint unspecified) | 62.8 | |
| 5 | DeepSeek V3.2 | 60.9 | #419 |
| 6 | Kimi K2.5 | 59.8 | #266 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=tebench-stale-test-update · How It Works · Data refreshed daily, snapshot 2026-11-21.