谷歌近日正式推出其最新的前沿人工智能模型 Gemini 4 Carbon，该模型在编码、代理任务以及网络安全领域展现出显著进步 12。此次发布距离上一代模型 Gemini 4 Argon 的问世仅间隔两周时间 2。尽管新模型技术能力强劲，但谷歌目前并未向公众开放使用权限，而是采取分阶段发布的策略 2。当前，Gemini 4 Carbon 仅向少数选定的网络安全防御公司提供服务 1。未来，谷歌计划通过 Fairwind Program 项目进一步扩大合作伙伴范围，使更多机构能够接入这一模型 1。
在基准测试表现方面，Gemini 4 Carbon 目前在 Artificial Analysis Intelligence Index (v5.0) 榜单中位居榜首，超越了竞争对手 Claude Opus 5.5 和 GPT-6 Astra 1。该模型在科学分析方面也达到了新的阈值水平 2。针对网络安全的特定应用场景，Gemini 4 Carbon 在漏洞检测、验证及修复环节取得了重大进展 2。数据显示，相比前代产品 Argon，Carbon 生成无漏洞代码的可能性高出两倍 2。此外，Chrome 团队指出，该模型生成高质量补丁的能力比非谷歌模型强出十倍 2。在 WebDev Arena 排行榜上，Gemini 4 Carbon 的得分增加了 89 点，排名第 4 位 2。
Google has officially announced the launch of its latest frontier AI model, Gemini 4 Carbon, marking a significant step forward in coding capabilities, agentic task execution, and cybersecurity. 1 The new model arrives just two weeks after the release of its predecessor, Gemini 4 Argon, but is currently not available for general public use. 2 Instead, access is restricted to a select group of cybersecurity defense companies, with plans to expand availability through the Fairwind Program in the future. 1
According to benchmark data, Gemini 4 Carbon has secured the top position on the Artificial Analysis Intelligence Index (v5.0), outperforming competitors such as Claude Opus 5.5 and GPT-6 Astra. 1 In terms of specific performance metrics, the model demonstrates a twofold increase in generating code without vulnerabilities compared to Argon. 2 Furthermore, the Chrome team reports that Carbon is ten times more effective at producing high-quality patches than non-Google models, while its score on the WebDev Arena leaderboard rose by 89 points to reach fourth place. 2 These improvements highlight substantial gains in detecting, validating, and repairing security flaws within software systems. 2
评论
还没有评论，欢迎留下第一条。