AI News 2026-08-26 — Top 10 AI Stories

Curated AI news digest for 2026-08-26 (timezone Asia/Shanghai).

  1. Jalapeño’s first results show industry-leading speed and efficiency in AI inference — OpenAI News. Jalapeño is a custom inference chip from OpenAI that delivers faster, more power-efficient AI inference, with higher throughput and lower latency for modern models.
  2. Claude Cowork finally remembers what you told the app in chat — TechCrunch AI. Anthropic is giving Claude a shared memory across chat and Cowork, so users no longer have to repeatedly brief the AI on projects, preferences, and other context.
  3. OpenAI loses a top data center exec, as stream of high-profile departures continues — TechCrunch AI. Before Malone left, OpenAI had already reshuffled its infrastructure org, shifting his reporting line away from President Greg Brockman and putting Vice President Sachin Katti in charge of the group.
  4. 基于阿里千问，路透社推出首个自研大模型 — 雷锋网. 8月25日，据外媒报道，国际知名新闻机构汤森路透（路透社母公司）近日推出了其自研大模型Thomson-1。该模型基于阿里Qwen大模型开发，旨在降低对Claude等闭源模型的依赖并减少长期使用成本。 此前，汤森路透曾推出自己的AI助手，接入的是海外大模型厂商Anthropic的Claude大模型。出于降低闭源模型的依赖和成本因素考量，近期，汤森路透选择基于阿里开源模型Qwen3.5自研出Thomson-1 大模型。Thomson-1 将首先用于表格分析、文档审核等专业工作，并逐步承接部分此前由 Claude 完成的任务。 开源模型允许企业自行部署和调整…
  5. OpenAI’s Jalapeño chip is built for fast inference at scale, benchmarks show — TechCrunch AI. Tested on SemiAnalysis’ InferenceX benchmark, Jalapeño registered both more tokens per user and more throughput per kilowatt than the currently available state-of-the art.
  6. The full stack behind abundant intelligence — OpenAI News. OpenAI CFO Sarah Friar explains how advances across chips, compute, models, and products compound to deliver more useful intelligence at greater scale and lower cost.
  7. OpenAI subpoenaed by Alabama AG over Hugging Face hack — The Verge AI. Alabama's attorney general issued a subpoena to OpenAI on Monday as part of an investigation into how one of its AI agents escaped a supposedly secure testing environment and autonomously hacked another company last month. The investigation seeks to determine whether OpenAI's sa…
  8. 通用模型竟然比医疗专用模型更懂医疗？一篇 ACL 论文的两个反直觉发现 | GAIR Paper 123 — 雷锋网. Doubao 、DeepSeek 在三语数据集上的表现，普遍超过 GPT-4o 和 Gemini。 作者丨幸丽娟 编辑丨岑 峰 医疗，是大模型落地最特殊的场景之一。 特殊在哪里？特殊在一个很微小的错误，就可能导致致命的后果；也特殊在你不能用直觉去判断它"行"还是"不行"—— 你觉得医疗专用模型肯定比通用模型更懂医疗？直觉上是。但实验数据不支持。 你觉得英语训练出来的模型处理英语医疗数据肯定最好？直觉上是。但实验数据也不支持。 这两重"反直觉"，正是纽约大学阿布扎比分校和阿布扎比克利夫兰诊所的研究团队在ACL 2026 Findings上发表的论文揭示的…
  9. 赚翻了！董宇辉等大主播出走后，东方甄选净利润反而暴涨86倍；字节将推统一办公品牌「豆包工作」；宇树跌至全球人形机器人市值榜第二 — 雷锋网. 要闻提示 1.赚翻了！董宇辉等大主播出走后，东方甄选薪酬少发4个亿，净利润暴涨86倍 2.新一代玄戒芯片正式发布，雷军透露小米重启大芯片研发累计投入研发经费超 210 亿 3.TRAE、扣子并入豆包，字节将推统一办公品牌“豆包工作” 4.宇树科技跌至全球人形机器人市值榜第二 5.任少卿创立具身公司，蔚来将成新公司战略股东 6.做多阿里巴巴！蔡崇信增持8000万港元，吴泳铭增持4000万港元 7.腾讯IEG重组AI游戏团队，或扩编至500人 8.Anthropic两款新Claude模型曝光 今日头条 赚翻了！董宇辉等大主播出走后，东方甄选薪酬少发4个亿，…
  10. 从开源走向共建：范式联合优必选等十余家具身巨头发布PhanthyMotus新计划 — 量子位. 近日，范式正式举办 PhanthyMotus 生态社区共建计划发布会，宣布其首个通用具身Agent底座从“开源”迈入“多方共建”新阶段。

Back to latest AI News