# Google previews Gemini 4 Carbon, available only to select security partners for now

> Original title: 谷歌又「预」发布了一个 Gemini 4 最强模型？

- Source: meng shao
- Published: 2026-10-29T02:46:44.000Z
- Bad Influence: https://badinfluence.news/items/vbs9sk0noq7629q3ivupd6q7u
- Original: https://x.com/shao__meng/status/2115636421637890425

## Summary

Google announced Gemini 4 Carbon, which it says leads the Artificial Analysis Intelligence Index v5.0 with a score of 59, ahead of Claude Opus 5.5 at 58. Access is limited at first to a network security defense group and Fairwind trusted partners, with introductory pricing of about $2 per million input tokens and $10 per million output tokens. The benchmark figures cited, including SEC-bench Pro at about 88.2%, are the company's own or partners' claims, and the author notes they need separate verification.

## Post

谷歌又「预」发布了一个 Gemini 4 最强模型？

Gemini 4 Carbon 在 Artificial Analysis Intelligence Index v5.0 第一（得分 59，Claude Opus 5.5 得分 58），在漏洞发现、利用验证、补丁修复等防御向基准上宣称领先；但现在还不能用，先给部分网络安全防御方与可信伙伴（Fairwind 计划），并参与美国政府自愿的发布前测试，再逐步放开给开发者、企业和消费者。
https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon

能力叙事：更难、更长的工程任务
Carbon 面向更难、跨时更长的工程问题，适合持续的 Agent 工作流。内部团队称它能跑上数天，在卡死之后仍继续推进，例如在大型代码库与底层系统里定位根因并打补丁。反馈驱动的改进覆盖编码、Agent、科学任务、全栈 Web 与多模态理解等。Web 开发方面，称输出质量提升后，WebDev Arena 上升约 +89 Elo，排到第 4（约 1768 分）。

网络安全
官方称在「发现 → 验证是否可利用 → 修复根因」整条链上有显著跃升；外部基准均为未加护栏评测口径：
· SEC-bench Pro 约 88.2%（在真实软件中自主挖复杂漏洞）
· ExploitGym 约 54.8%（判断漏洞是否可被利用，便于排优先级）
· Collinear CWE-bench 约 71.0%（把已验证发现变成可工作的补丁）

对内指标还包括：相对 Argon，生成安全代码成功率自称提升超过约 2 倍；Chrome 团队称其高质量补丁表现比「另一家实验室的大型前沿模型」强约 10 倍以上；云安全公司 Wiz 也称其在 Web 渗透测试相关基准上领先（与 Wiz 的 Scan for Good 相关）。这些多为合作方或内部观察，口径需单独核实。

安全与对齐
沿用 Argon 的多层安全栈（训练、实时滥用防护、提示注入韧性、思维链监控等），并称在对齐上再推进一步，尤其是高压 / 复杂任务上「该怎么做就怎么做」：
· Terminal-Bench 4.0 上，成功作弊降到 0.0%，作弊尝试约 0.4%，同时整体正确率仍提升；内部现场编码里，高严重度错位（misalignment）告警约降 80%。
· 相对 Argon，无依据断言约减半；更常在交付物里写方法学限定，或明确列出缺什么证据、为何暂不执行。
遇歧义时更倾向先拉来源、把矛盾亮给用户再动手。
· 越权约降 26%、服务不足约降 29%，多步计划更易做完且不越界。
· Gray Swan 间接提示注入（IPI）基准：Argon 攻击成功率约 0.7%，Carbon 称 0.0%。

发布与价格
本周起向一部分网络安全防御方开放，再经 Fairwind 逐步扩到更多可信伙伴；收集反馈、迭代护栏后，再尽快上产品面。介绍期定价与 Argon 相同：输入约 $2 / 百万 token，输出约 $10 / 百万 token，缓存输入约按输入价 95% 折扣。介绍期结束后，将变为输入约 $4、输出约 $20（每百万 token）。

## AI translation

Did Google "preview" the strongest Gemini 4 model again?

Gemini 4 Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 (score 59, versus 58 for Claude Opus 5.5), and claims leads on defense-oriented benchmarks for vulnerability discovery, exploit verification, and patch fixing. But it isn't usable yet: it's first being given to some cybersecurity defenders and trusted partners (the Fairwind program), and is taking part in voluntary pre-release testing with the US government, before being gradually opened to developers, enterprises, and consumers.
https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon

Capability narrative: harder, longer engineering tasks
Carbon targets harder engineering problems that span longer periods, suited to sustained Agent workflows. Internal teams say it can run for days, keep pushing forward after getting stuck, for example locating root causes and writing patches in large codebases and low-level systems. Feedback-driven improvements cover coding, agents, scientific tasks, full-stack web, and multimodal understanding. On web development, it is said that after improved output quality, it rose about +89 Elo on WebDev Arena, reaching 4th place (about 1768 points).

Cybersecurity
Officially, there is a significant leap across the whole chain of "discovery → verifying exploitability → fixing the root cause"; external benchmarks are all evaluated without added guardrails:
· SEC-bench Pro about 88.2% (autonomously finding complex vulnerabilities in real software)
· ExploitGym about 54.8% (judging whether a vulnerability can be exploited, to help prioritize)
· Collinear CWE-bench about 71.0% (turning verified findings into working patches)

Internal metrics also include: compared with Argon, the success rate for generating secure code is claimed to have improved by more than about 2x; the Chrome team says its high-quality patch performance is more than about 10x stronger than "a large frontier model from another lab"; cloud security company Wiz also says it leads on benchmarks related to web penetration testing (tied to Wiz's Scan for Good). These are mostly partner or internal observations, and the claims need to be verified separately.

Safety and alignment
It keeps Argon's multi-layer safety stack (training, real-time abuse protection, prompt injection resilience, chain-of-thought monitoring, etc.), and claims to have taken alignment a step further, especially on high-pressure or complex tasks, doing "what should be done":
· On Terminal-Bench 4.0, successful cheating dropped to 0.0%, with cheating attempts at about 0.4%, while overall accuracy still improved; in internal live coding, high-severity misalignment alerts dropped about 80%.
· Compared with Argon, unsupported assertions are roughly halved; it more often adds methodological qualifiers to its deliverables, or explicitly lists what evidence is missing and why it hasn't acted yet.
When faced with ambiguity, it leans toward first pulling sources and surfacing contradictions to the user before acting.
· Overreach dropped about 26% and under-service about 29%, making multi-step plans more likely to be completed without exceeding scope.
· On the Gray Swan indirect prompt injection (IPI) benchmark: Argon's attack success rate was about 0.7%, and Carbon is claimed at 0.0%.

Release and pricing
Starting this week, it opens to a subset of cybersecurity defenders, then expands gradually to more trusted partners through Fairwind; after collecting feedback and iterating on guardrails, it will reach product surfaces as soon as possible. The introductory pricing is the same as Argon: about $2 per million input tokens, about $10 per million output tokens, with cached input at about a 95% discount off the input price. After the introductory period ends, it will become about $4 input and $20 output (per million tokens).

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHVxAAv6bAAAU8RF.jpg&mode=thumb&exp=1793491200&sig=ec937df80a7475e4)

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHVxADK3boAE68AW.jpg&mode=thumb&exp=1793491200&sig=2d6145a195e6fdbb)

## Quoted post @GoogleDeepMind

> Gemini 4 Carbon is our next frontier model.
> 
> It’s state-of-the-art across cybersecurity and long-horizon capabilities.

https://x.com/GoogleDeepMind/status/2115489356437651654