Google发布Gemini 4 Carbon，安全与长程能力领先
🤖 AI Agents Weekly: Gemini 4 Carbon, GPT-6.1 Terra, OpenAI's Three Discoveries, Mistral Large 4, Axiom CFSG Proof, Qwen Investigation, Genie 4, and More
全新前沿模型发布，安全与长程能力数据详实，直接对标竞品，值得从业者关注其实际落地表现。
In today’s issue:
本期内容：
- Google previews Gemini 4 Carbon
- OpenAI ships GPT-6.1 Terra
- OpenAI models refute Exponential Time Hypothesis
- Mistral opens Large 4 weights
- Agent swarm formalizes 15,000-page proof
- Cohere flags Qwen upload attempt
- Anthropic measures agent influence on experts
- Cognition launches FrontierCode 2.0
- DeepMind unveils Genie 4
- Pretraining poison backdoors CLI agents
- Google 预览 Gemini 4 Carbon
- OpenAI 发布 GPT-6.1 Terra
- OpenAI 模型反驳指数时间假设
- Mistral 开放 Large 4 权重
- 智能体群形式化验证 15,000 页证明
- Cohere 标记 Qwen 上传尝试
- Anthropic 衡量智能体对专家的影响
- Cognition 推出 FrontierCode 2.0
- DeepMind 发布 Genie 4
- 预训练投毒后门攻击 CLI 智能体
And all the top AI dev news, papers, and tools.
以及所有顶级 AI 开发新闻、论文和工具。
Top Stories
头条新闻
Gemini 4 Carbon
Google DeepMind introduced Gemini 4 Carbon, its next frontier model, with state-of-the-art cybersecurity and long-horizon capabilities. Google says internal teams have seen it run for days on multi-step engineering work such as root-causing and patching bugs in large codebases.
Google DeepMind 推出了其下一代前沿模型 Gemini 4 Carbon，具备最先进的网络安全和长周期能力。Google 表示，内部团队已观察到它在大型代码库的根因分析和漏洞修补等多步工程任务中连续运行数天。
- Benchmarks: Carbon ranks first on the Artificial Analysis Intelligence Index v5 at 59, ahead of Claude Opus 5.5 (58) and GPT-6 Astra (57), and Google’s comparison table shows 65.2% on Terminal-Bench 4.0 and 77.4% on Agents’ Last Exam v1.1, ahead of GPT-6 Astra and Claude Opus 5.5. Cognition’s FrontierCode 2.0 puts it at 40%, behind Claude Opus 5.5, GPT-6 Astra, and Claude Sonnet 5.5.
- Cybersecurity: Evaluated without guardrails, it reaches 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.
- Alignment: Google reports successful cheating on Terminal-Bench 4.0 at 0.0%, almost 80% fewer high-severity misalignment alerts in internal live coding environments, and a 0.0% attack success rate on Gray Swan prompt injection tests.
- Cost: Artificial Analysis measures about $7.76 per Index task, roughly 23% less than Claude Opus 5.5 at max effort, with a 1M-token context window.
- Availability: Access starts with a subset of cyber defenders through the Fairwind Program, and Google says developers, businesses, and consumers will follow. Pricing is $2 input and $10 output per million tokens during the introductory period, then $4 and $20.
- 基准测试：Carbon 在 Artificial Analysis Intelligence Index v5 中以 59 分位居第一，领先于 Claude Opus 5.5（58 分）和 GPT-6 Astra（57 分）。Google 的对比表显示其在 Terminal-Bench 4.0 上得分为 65.2%，在 Agents’ Last Exam v1.1 上得分为 77.4%，均领先于 GPT-6 Astra 和 Claude Opus 5.5。Cognition 的 FrontierCode 2.0 将其排名为 40%，落后于 Claude Opus 5.5、GPT-6 Astra 和 Claude Sonnet 5.5。
- 网络安全：在无护栏评估下，它在 SEC-bench Pro 漏洞发现任务中达到 88.2% 的得分，在 ExploitGym 利用任务中达到 54.8%，在 CWE-bench 修补任务中达到 71.0%。
- 对齐性：Google 报告称在 Terminal-Bench 4.0 作弊成功率为 0.0%，内部实时编码环境中的高严重性不对齐警报减少了近 80%，在 Gray Swan 提示注入测试中的攻击成功率也为 0.0%。
- 成本：Artificial Analysis 测量每个 Index 任务成本约为 7.76 美元，比 Claude Opus 5.5 在最大努力下的成本低约 23%，上下文窗口为 100 万 token。
- 可用性：访问权限首先通过 Fairwind Program 向部分网络防御者开放，Google 表示开发者、企业和消费者将随后获得访问权限。在入门期间，每百万 token 输入价格为 2 美元，输出价格为 10 美元；之后分别为 4 美元和 20 美元。
Blog | Artificial Analysis
博客 | Artificial Analysis
Read more
阅读全文
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力