AI Tracker下載 App

基礎架構新聞，
為開發者整理。

各家 AI 公司最新的基礎架構動態：發表、調價和停用，並附上對你的影響。

OpenAI· 基礎架構

OpenAI 說明如何在 Azure Cosmos DB 上擴展其儲存平台 Habitat

在這篇日期為 2026 年 10 月 23 日的工程系列第二部分中，OpenAI 介紹了 Habitat——支援 ChatGPT、API 和 Codex 的線上儲存平台。OpenAI 表示，Habitat 為超過 40 個區域的 10 億多名使用者提供服務，儲存超過 500 PB 的資料；OpenAI 對文章的摘要稱，ChatGPT 的讀取速度提升了 12 倍。程式開發

Polylane 推出主動式代理，監控並修復 Cloudflare Workers 應用程式

Boris Tane 宣布推出 Polylane for Cloudflare，這項服務會執行主動式代理，持續監控 Cloudflare 帳戶中的 Workers、Durable Objects、Workflows 等資源。根據公告，這些代理會修復錯誤、提升效能並處理警示。基礎架構

Vercel 說明 Vercel Agent 如何取得正式環境的存取權

Vercel 發布了一篇文章，說明 Vercel Agent 如何存取正式環境部署。Vercel 表示，Agent 會以唯讀權限調查問題，接著提出修復計畫以及執行修復所需的權限，因此要求它修復部署並不代表要交給它廣泛的憑證。基礎架構

Vercel AI Gateway 新增 tako_search，提供結構化資料搜尋工具

Vercel 宣布，Tako 的結構化資料搜尋現已作為工具在 AI Gateway 上提供。這項工具不會回傳一大堆文字頁面，而是以 JSON 格式提供包含單位、來源和日期的資料：Vercel 列出的資料涵蓋股票、民調、天氣和政府合約，並表示這些資料均已取得授權、附有引用來源且保持最新。Anthropic· 基礎架構

Claude Models on Google Cloud（Anthropic 管理）現已正式推出

Anthropic 表示，Claude Models on Google Cloud（Anthropic 管理）現已正式推出。據 Anthropic 說明，此服務將完整的 Claude Platform 功能帶到 Google Cloud，並支援 Google Cloud IAM 憑證、帳單，以及抵扣承諾用量。基礎架構

Vercel Drop 可免帳號部署檔案、資料夾或 ZIP

Vercel 宣布推出 Vercel Drop：使用者拖放檔案、資料夾或 ZIP，即可取得可供瀏覽的網址，無須帳號。只有在一小時內登入認領，專案才會持續上線。基礎架構

You.com 推出 Inference Router：提供內建網頁搜尋、相容於 OpenAI API 的開放權重模型 API

You.com 執行長 Richard Socher 宣布推出 Inference Router，這是一項託管推論服務，提供單一相容於 OpenAI API 的介面，且不收取平台費。現有的 OpenAI 相容程式碼只需更改 base URL 即可切換。Google· 基礎架構

Google Cloud announces Agent Compute Engine and expanded agent governance for Gemini for business

Google Cloud announced a new product, Agent Compute Engine, at AGNTCon + MCPCon. It also announced expanded agent governance for Gemini for business. Google Cloud says the two together give organizations the control they need to run, govern and scale their agents. The details are in Google Cloud's blog post, "Introducing Agent Compute Engine and expanding agent governance".Anthropic· 程式開發

Anthropic releases Claude API version 2026-10-21, with breaking changes

Anthropic's ClaudeDevs account announced API version 2026-10-21, which it calls the first official Claude API version release since 2023. Anthropic says it plans to release versions more often, so developers can handle breaking changes before a new model requires them. The older version, 2023-06-01, is not deprecated.OpenAI· 基礎架構· 1 則討論

OpenAI will require approval for model routers and AI gateways that offer its models

OpenAI says it has spent the past few months aligning safety policies with the partners and customers that provide its models, whether they host them or resell access. Partners must maintain safeguards consistent with OpenAI's own, investigate potential violations, take action and coordinate enforcement with OpenAI.OpenAI· 基礎架構

OpenAI API lets organizations choose which service tiers projects can use

OpenAI's API changelog for October 21, 2026 says that Platform settings now let you choose which service tiers projects can use, and that organization settings can restrict service tiers for all projects. The entry points to documentation on service tier errors.OpenAI· 基礎架構

OpenAI describes how ChatGPT's ad stack works

OpenAI published a post, summarized by @OpenAIDevs on X on October 21, on how ChatGPT's advertising system is built. OpenAI says it is designed to respond to the evolving conversation while keeping answers independent of ads.基礎架構

Vercel makes WebSocket support in Vercel Functions generally available

Vercel announced that WebSocket support in Vercel Functions is now generally available. Vercel says it now serves 15 million WebSocket connections a day across AI chat, streaming and realtime multiplayer apps. The details are in Vercel's changelog entry.提醒OpenAI· 模型· 5 則討論

OpenAI says its models helped find critical security flaws in PHP

OpenAI announced that its models helped researchers, working under its "Patch the Planet" initiative, discover critical security flaws in PHP, which it says runs an estimated 70% of websites. According to OpenAI, it led a coalition of maintainers, researchers and security providers for about two months to prepare fixes and protections before disclosure; the fix was intricate and required a major change to the PHP engine.OpenAI· 模型· 4 則討論

OpenAI says its models are coming to Google Cloud

OpenAI has published a post titled "OpenAI models are coming to Google Cloud". Its summary says OpenAI and Google Cloud are bringing frontier models and Codex workflows to Gemini's Model Garden, available through Google Cloud Marketplace. This is OpenAI's own account. Posts on X about the announcement say the models and Codex arrive on Google Cloud in the coming weeks.DeepSeek· 模型

Stork: Alibaba raised its price for DeepSeek V4 Flash 0731 by 230%

Stork AI Daily, a newsletter that runs its own LLM price catalog, reports that Alibaba raised the price of DeepSeek-V4-Flash-0731 overnight. This is Stork's account from its own tracker; neither Alibaba nor DeepSeek has announced it, and the newsletter does not say which Alibaba service it means.程式開發· 1 則討論

Vercel introduces durable-harness for running coding-agent harnesses durably

Vercel announced durable-harness on X. In Vercel's words, it lets users run Pi, OpenCode, Codex, Claude Code, eve, fx or their own harness durably on Vercel, and agents automatically resume work after failures and run for hours, days or months.OpenAI· 基礎架構

OpenAI API adds a 12-hour prompt caching option

OpenAI's API changelog for October 20, 2026 says developers can now choose 12-hour prompt caching in the API (labelled for v1/responses). It keeps shared instructions, tools and context cached longer, so cached-input pricing applies even when requests are hours apart. OpenAI says 12-hour cache writes cost 2.5x the input token price.DeepSeek· 基礎架構

DeepSeek open-sources DeepInfer, its inference engine

DeepSeek has published DeepInfer on GitHub (deepseek-ai/DeepInfer) under the Apache-2.0 licence. The README calls it the inference framework that powers DeepSeek's large-scale production services. The release was shared on Hacker News on October 20. The details below are from the repository README.GitHub Copilot· 基礎架構

GitHub adds a REST API for Copilot model policies

GitHub's changelog announces generally available GET and PATCH REST endpoints to read and update Copilot model policies at four levels: enterprise, enterprise teams, enterprise apps and organizations (for example /orgs/{org}/copilot/model-policies).
較早的新聞