PROBE (Code Generation) - Java: leaderboard
The Java track of PROBE (UNC Charlotte and University of Coimbra, Empirical Software Engineering 2026): models write complete Java programs for 1,651 English-language Project CodeNet problems with extended unit tests, five samples per problem under the baseline prompt. The score is the passed outcome rate, the share of unit-test executions that pass, in percent; higher is better.
Metric: Passed outcome rate (%): share of unit-test executions that pass, averaged over the unit tests of each problem, 5 generated samples per problem and the 1,651 filtered Project CodeNet problems, for complete Java programs under the baseline prompt; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 Mini | 69 | #779 |
| 2 | Gemini 2.0 Flash | 59 | #704 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=probe-code-generation-java · How It Works · Data refreshed daily, snapshot 2026-11-13.