Expired50% confidenceFactExact time
Sonar evaluated 53+ large language models using 4,444 Java programming tasks to assess code quality
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Sonar Tests Code Quality Across 53 LLMs: Claude Has the Most Security Vulnerabilities, GPT-5 Code Volume Surges 5x
bilibiliAI_Express6/1/2026
Related Claims
Unverified大语言模型编程场景的核心能力指标包括HumanEval代码生成基准、SWE-bench真实Bug修复成功率及跨文件依赖解析准确性68% similarUnverifiedLarge language models' performance in code generation is highly dependent on the scale and quality of pre-training data.68% similarUnverifiedLarge language models perform better on code explanation tasks than on code generation tasks.65% similarUnverified大语言模型处理长任务的能力受制于上下文窗口的大小与质量64% similarUnverifiedThe current mainstream large language models' understanding of code is essentially statistical pattern matching based on massive code corpora, lacking the genuine ability to model program runtime states.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/50491API
curl https://kongchang.com/api/v1/knowledge/claims/50491MCP
get_claim(id=50491)