官方更新 DeepMind Blog Wed, 28 Oct 2026 正文 ok 254 words

Gemini 4 Carbon:谷歌最新前沿模型发布,具备顶尖网络安全与长程任务能力

英文标题:Gemini 4 Carbon: Our next frontier model

谷歌正式宣布推出新一代前沿模型 Gemini 4 Carbon,主打业界顶尖的网络安全防御与长程任务(long horizon)处理能力。在最新更新的 Artificial Analysis Intelligence Index v5.0 榜单中,Carbon 以 59 分位列第一,并在 SEC-bench Pro(88.2%)、ExploitGym(54.8%)和 CWE-bench(71.0%)等网络安全与代码基准测试中取得全面领先。为确保安全,谷歌正通过美国政府预发布测试及 Fairwind 项目向特定防御专家和合作伙伴分阶段开放,后续将推向开发者与企业。该模型首发价格与 Argon 一致,输入每百万 tokens 2 美元,输出每百万 tokens 10 美元,缓存输入享受 95% 折扣。
打开原文相关 2.4 · 机会 1
media9.8 KB media1.6 KB media75.6 KB media143.4 KB media18.3 KB media35.0 KB media23.9 KB media45.2 KB media27.4 KB media52.2 KB media28.1 KB media53.3 KB media3.0 KB media8.8 KB media3.6 KB media9.8 KB media2.8 KB media8.3 KB media4.1 KB media9.2 KB

中文

今天,我们正式发布下一代前沿模型:Gemini 4 Carbon。这是我们迄今为止最强大的模型,具备业界领先(SOTA)的网络安全能力和长程任务(long horizon)处理能力。

在最新更新的 Artificial Analysis Intelligence Index v5.0 中,Carbon 以 59 分位居榜首。它在编程任务、Agent 任务以及对网络安全防御至关重要的核心能力上均实现了大幅提升,在漏洞挖掘基准 SEC-bench Pro(88.2%)、漏洞利用基准 ExploitGym(54.8%)以及漏洞修复基准 CWE-bench(71.0%)中均确立了稳固的领先优势。在内部基准测试中,Carbon 大幅提升了安全代码的生成表现;在与我们自身团队的实时编程环境中,它也展现出更低的不对齐率与内容捏造率。

继两周前推出的 Gemini 4 Argon 之后,Carbon 再次印证了我们向前推展技术前沿的速度。我们的研究人员与模型紧密协作,在加速能力突破的同时,同步强化各项安全防护机制。

我们正在采取审慎严谨的方式推进部署,并积极参与美国政府面向模型发布前测试的自愿流程。从本周开始,我们将向部分网络安全防御专家开放 Carbon,并随后通过 Fairwind Program 逐步分阶段向更多受信合作伙伴提供访问权限。在尽快通过我们的产品界面向开发者、企业及广大消费者开放 Carbon 之前,我们将持续收集反馈并迭代安全护栏。

Gemini 4 Carbon 的首发价格将与 Argon 保持一致:输入价格为每 100 万 tokens 2 美元,输出价格为每 100 万 tokens 10 美元,缓存输入 tokens 价格在基础输入价格上享受 95% 折扣。

English / 原文

Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59. It shows significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense, landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching. On internal benchmarks, Carbon drives a strong uplift in generating secure code, and in live coding environments with our own teams it shows reduced misalignment and fabrication rates.

Building on Gemini 4 Argon, which was launched two weeks ago, Carbon illustrates the pace at which we’re pushing the frontier forward. Our researchers and models are working hand in hand to advance capabilities faster, while strengthening our safeguards in tandem.

We're taking a deliberate approach to rollout and are actively engaged in the U.S. government’s voluntary process for pre-release model testing. Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program. We’re continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible.

Gemini 4 Carbon will launch at the same introductory price as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.