AI Weekly: GPT-5.6-Cyber, Muse Glimmer, and the Agent Browser
AI Weekly: GPT-5.6-Cyber, Muse Glimmer, and the Agent Browser
AI 周报：GPT-5.6-Cyber、Muse Glimmer 与智能体浏览器
Week of August 5 to August 12, 2026. This Week at a Glance: OpenAI shipped GPT-5.6-Cyber on August 10, a purpose-trained security model behind its Daybreak Red approval gate, priced at $12.50 per million input tokens and $75 per million output. 2026 年 8 月 5 日至 8 月 12 日。本周概览：OpenAI 于 8 月 10 日发布了 GPT-5.6-Cyber，这是一款置于其 Daybreak Red 审批门槛之后的专用安全模型，输入价格为每百万 token 12.50 美元，输出价格为每百万 token 75 美元。
Meta returned to open weights with Muse Glimmer on August 10, a 30-billion-parameter Apache 2.0 model that runs on a single 24GB consumer GPU and targets local agent workflows. ByteDance released Seedance 2.5 on August 8, and Alibaba shipped Qwen3.8-Max on August 3, keeping the release calendar full outside the two headline drops. Meta 于 8 月 10 日通过 Muse Glimmer 回归开源权重阵营，这是一个拥有 300 亿参数的 Apache 2.0 模型，可在单张 24GB 消费级显卡上运行，旨在支持本地智能体工作流。字节跳动于 8 月 8 日发布了 Seedance 2.5，阿里巴巴于 8 月 3 日发布了 Qwen3.8-Max，使得除了上述两个重磅发布外，本周的发布日程依然排得满满当当。
OpenAI’s Codex added forkable thread history, Amazon Bedrock login, audio inputs, and imports from Cursor and Claude Code settings, tightening the agentic coding race. Cursor rolled out Cursor Router with Auto Intelligence and Auto Balance, claiming above-Fable satisfaction at 68 percent lower cost. OpenAI 的 Codex 增加了可分叉的线程历史记录、Amazon Bedrock 登录、音频输入以及从 Cursor 和 Claude Code 设置导入的功能，进一步加剧了智能体编程领域的竞争。Cursor 推出了带有自动智能（Auto Intelligence）和自动平衡（Auto Balance）功能的 Cursor Router，声称在成本降低 68% 的情况下，用户满意度超过了 Fable。
The MCP 2026-07-28 stateless specification is now the live standard, removing protocol-level sessions and the session-id header so any server instance can answer any request. Cloudflare launched Kitesurf on August 6, an agent-first browser that runs on Workers in V8 isolates and uses 3 to 7 times less CPU and memory than Chromium. MCP 2026-07-28 无状态规范现已成为正式标准，移除了协议层面的会话和 session-id 标头，使得任何服务器实例都能响应任何请求。Cloudflare 于 8 月 6 日推出了 Kitesurf，这是一款以智能体为先的浏览器，运行在 V8 隔离环境的 Workers 上，其 CPU 和内存占用比 Chromium 低 3 到 7 倍。
2027 DRAM and HBM capacity is reportedly sold out, with buyers receiving 60 to 70 percent of requested volumes and paying deposits upfront. Two releases defined the week, and they point in opposite directions. OpenAI narrowed access with a gated cyber model for approved defenders. Meta widened it with an open-weight model built to run on a laptop. 据报道，2027 年的 DRAM 和 HBM 产能已售罄，买家仅能获得需求量的 60% 到 70%，且需预付定金。本周的两个发布定义了行业走向，且指向截然相反的方向：OpenAI 通过为获批防御者提供的受限网络安全模型收窄了访问权限；Meta 则通过构建可在笔记本电脑上运行的开源权重模型扩大了访问范围。
The tooling, standards, and infrastructure news underneath both moves tells the same story: the industry is building the plumbing for agents that act, not just chatbots that answer. 这两项举措背后的工具、标准和基础设施新闻传达了同一个信息：整个行业正在为“能够行动的智能体”构建底层架构，而不仅仅是构建“能够回答问题的聊天机器人”。
Models: OpenAI Gates Cyber, Meta Opens the Laptop
模型：OpenAI 限制网络安全模型，Meta 拥抱笔记本电脑
OpenAI released GPT-5.6-Cyber on August 10, and the framing matters as much as the model. OpenAI’s documents list pricing for GPT-5.6-Cyber at $12.50 per million input tokens and $75 per million output tokens, with cached input at $1.25 per million tokens. That makes it the priciest member of the GPT-5.6 family by a wide margin. OpenAI 于 8 月 10 日发布了 GPT-5.6-Cyber，其发布框架与模型本身同样重要。OpenAI 的文档显示，GPT-5.6-Cyber 的定价为每百万输入 token 12.50 美元，每百万输出 token 75 美元，缓存输入为每百万 token 1.25 美元。这使其成为 GPT-5.6 家族中价格最高的一员，且溢价幅度巨大。
Sol, the flagship, lists at $5 per million input tokens and $30 per million output tokens for short-context use. The model is not for general use. It is an alias for OpenAI’s most advanced purpose-trained cybersecurity models, for approved defenders conducting authorized vulnerability research, exploit validation, and security testing, and it requires separate approval and provisioning through the Daybreak program. 旗舰模型 Sol 在短上下文使用场景下的定价为每百万输入 token 5 美元，输出 30 美元。该模型并非供通用使用，它是 OpenAI 最先进的专用网络安全模型的别名，旨在供获批的防御者进行授权的漏洞研究、漏洞利用验证和安全测试，且需要通过 Daybreak 项目进行单独审批和配置。
The gate is the product. Daybreak Red is for approved security teams doing advanced, authorized cyber work, including vulnerability research, penetration testing, red-team exercises, and exploit validation on systems the organization owns or has permission to test. The launch answers a specific complaint. Security engineers using OpenAI’s Codex Security product hit constant refusals on defensive work, because the general models find a bug and then decline to discuss it. GPT-5.6-Cyber reduces those refusals for vetted users. 门槛本身就是产品。Daybreak Red 专为获批的安全团队设计，用于执行高级、授权的网络安全工作，包括对组织拥有或获准测试的系统进行漏洞研究、渗透测试、红队演习和漏洞利用验证。此次发布回应了一个具体的抱怨：使用 OpenAI Codex Security 产品的安全工程师在进行防御性工作时经常遭遇拒绝，因为通用模型在发现漏洞后会拒绝讨论它。GPT-5.6-Cyber 为经过审查的用户减少了这些拒绝。
The pricing tells the safety story
定价背后的安全逻辑
The pricing structure on GPT-5.6-Cyber is a policy statement wearing a price tag. At $12.50 per million input tokens and $75 per million output, the model costs roughly 2.5 times Sol on input and 2.5 times on output, before the long-context multiplier. That premium is not about compute. It is about signaling that this capability is for serious, funded, professional security work, and pricing casual experimentation out of reach. GPT-5.6-Cyber 的定价结构是一份贴着价格标签的政策声明。在不考虑长上下文乘数的情况下，该模型的输入价格约为 Sol 的 2.5 倍，输出价格也是 2.5 倍。这种溢价并非源于计算成本，而是为了表明该能力仅供严肃、有资金支持的专业安全工作使用，并将随意的实验性使用排除在外。
Prompt caching and the token-efficiency angle
Prompt 缓存与 Token 效率视角
One under-covered detail from the GPT-5.6 family carries through to the cyber model: prompt caching changes. GPT-5.6 introduced more predictable prompt caching with explicit cache breakpoints and a 30-minute minimum cache life, and for GPT-5.6 and later models cache writes bill at 1.25x the uncached input rate while cache reads keep the 90 percent cached-input discount. GPT-5.6 家族中一个未被充分报道的细节延续到了该网络安全模型中：Prompt 缓存的变化。GPT-5.6 引入了更可预测的 Prompt 缓存，具有明确的缓存断点和 30 分钟的最低缓存寿命。对于 GPT-5.6 及后续模型，缓存写入按未缓存输入费率的 1.25 倍计费，而缓存读取则保持 90% 的缓存输入折扣。
For agent workloads that reuse long system prompts and tool definitions across many calls, that 90 percent read discount is where real money gets saved. A security agent scanning a large codebase reuses the same context repeatedly, and caching turns what would be a punishing bill into a manageable one. 对于在多次调用中重复使用长系统提示词和工具定义的智能体工作负载而言，这 90% 的读取折扣才是真正节省成本的地方。一个扫描大型代码库的安全智能体会反复使用相同的上下文，而缓存将原本高昂的账单变成了可控的支出。