Claude 3.5 Sonnet 在 SWE-bench Verified 上刷新纪录
Raising the bar on SWE-bench Verified with Claude 3.5 Sonnet Jan 06, 2025
Claude 3.5 Sonnet 在 SWE-bench 上的突破意味着 AI 编程助手离真正解决复杂工程问题更近了一步，做软件开发的团队可以关注这一能力提升对日常代码修复和功能开发的潜在影响。
Anthropic 宣布其 Claude 3.5 Sonnet 模型在 SWE-bench Verified 基准测试中取得了 49.
7% 的通过率，较此前最佳成绩提升了约 10 个百分点。该测试评估 AI 模型解决真实 GitHub 问题的能力，包括代码修复、功能实现等。
Claude 3.5 Sonnet 在多个类别中表现优异，尤其在需要多步推理和上下文理解的复杂任务上。
这一进展表明 AI 在软件工程自动化领域正快速接近人类水平。
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Claude 3.5 Sonnet, SWE-bench, 代码修复, 基准测试.
Primary source verification: published directly by Anthropic: Engineering with first-party authority. Cross-source corroboration: 11 independent sources have covered this development (including Claude: Blog, arXiv: OpenAI, 宝玉), confirming industry-wide signal strength.
