Sonar LLM Leaderboard - Java - Bug Density: leaderboard
SonarQube bugs (reliability issues) per 1,000 lines of generated Java code on Sonar's LLM Leaderboard, which has each model solve a fixed set of Java programming assignments and analyses the generated code with SonarQube; lower is better.
Metric: Bugs per 1,000 lines of code (lower is better). Source: www.sonarsource.com. Saturation forecast: Around July 2027. 74 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.6 | 0.52 | #113 |
| 2 | GPT-5.5 (Medium) | 0.52 | #33 (GPT-5.5) |
| 3 | GPT-5.1 Codex (High) | 0.52 | #168 (GPT-5.1 Codex) |
| 4 | GPT-5.5 (Low) | 0.55 | #33 (GPT-5.5) |
| 5 | Gemini 4 Argon (High) | 0.56 | #9 (Gemini 4 Argon) |
| 6 | GPT-6 Astra (Medium) | 0.56 | #6 (GPT-6 Astra) |
| 7 | GPT-6 Sol (High) | 0.59 | #19 (GPT-6 Sol) |
| 8 | Claude Opus 5 (Thinking) | 0.59 | #16 (Claude Opus 5) |
| 9 | Gemini 3 Pro (High) | 0.6 | #82 (Gemini 3 Pro) |
| 10 | Qwen 3.5 397B A17B (Thinking) | 0.61 | #154 (Qwen 3.5 397B A17B) |
| 11 | GPT-6.1 Sol (Medium) | 0.61 | #7 (GPT-6.1 Sol) |
| 12 | MiniMax-M2.7 | 0.62 | #270 |
| 13 | GPT-5.1 (High) | 0.62 | #145 (GPT-5.1) |
| 14 | GLM-5.2 | 0.63 | #80 |
| 15 | Gemini 3.1 Pro (Preview) (High) | 0.63 | #64 (Gemini 3.1 Pro (Preview)) |
No result here: #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=sonar-llm-leaderboard-java-bug-density · How It Works · Data refreshed daily, snapshot 2026-11-05.