Gemini 3.7 Flash 发布
英文标题:Introducing Gemini 3.7 Flash
Gemini 3.7 Flash 在调试、问题解决和生产级代码生成等编程任务上,相比 3.6 Flash 有显著提升。在 FrontierCode 1.1 Main 和 DeepSWE v1.1 基准中,其得分分别从 34.4% 提升至 43.6%、从 49.0% 提升至 65.3%。该模型也提升了 Web 开发、复杂文档理解及真实业务工作流自动化能力,对需要构建应用、处理专业领域知识和自动化业务流程的开发者具有更高实用价值。
中文
3.7 Flash 在调试、问题解决等编程任务上较 3.6 Flash 实现显著提升。它还具备更高的首轮代码准确率,并且在生成生产就绪代码方面表现更佳,体现在 FrontierCode 1.1 Main(43.6% vs 34.4%)和 DeepSWE v1.1(65.3% vs 49.0%)基准测试中。
在 Web 开发方面,3.7 Flash 能够以更少的提示生成更具功能性的页面布局和功能更完整的应用。对于 UI 生成,无论参考输入是截图、图片还是完整设计系统,该模型都展现出较高的设计遵循度和一致性。它在 Arena.ai 的 WebDev Arena 中优于 3.6 Flash,Elo 评分为 1588,而 3.6 Flash 为 1538。
对于金融、法律和生物科学等知识密集型领域,3.7 Flash 提供了更强的推理能力和准确性。在用于测试模型处理复杂文档能力的 GDP.pdf 基准中,它显著优于 3.6 Flash(34.0% vs 22.0%)。它还在 AutomationBench 中超过 3.6 Flash,表明其能够更有效地完成真实世界的业务工作流(30.4% vs 17.0%)。
English / 原文
3.7 Flash shows strong gains over 3.6 Flash in coding tasks like debugging and issue resolution. It also achieves higher first-pass code accuracy and has improved performance in generating production-ready code as seen in FrontierCode 1.1 Main (43.6% vs 34.4%) and DeepSWE v1.1 (65.3% vs 49.0%).
In web development, 3.7 Flash generates more functional layouts and feature-complete apps in fewer prompts. For UI generation, the model shows high design adherence and parity based on a reference input, whether it’s a screenshot, an image, or a full design system. It outperforms 3.6 Flash on Arena.ai’s WebDev Arena with an Elo score of 1588 vs 1538.
For knowledge-dense fields like finance, law, and biosciences, 3.7 Flash delivers improved reasoning and accuracy. It significantly outperforms 3.6 Flash on the GDP.pdf benchmark (34.0% vs 22.0%), an eval for testing a model’s ability to process complex documents. It also surpasses 3.6 Flash in AutomationBench, demonstrating it can more effectively complete real-world business workflows (30.4% vs 17.0%).