跳到正文
9月9日周三
  1. Claude:Blog67

    Anthropic 发布 Claude Platform 降本指南并更新 claude-api 技能

    Anthropic 介绍用 Claude Platform 降低成本、保持性能的三种方法:提高提示词缓存命中率、升级到前沿模型时清除提示词反模式、按任务校准 effort。

    推荐理由:官方给出了提示词缓存、反模式清理和 effort 校准三类降本方法及实测数字,可直接迁移到自己的 Claude API 应用。

9月8日周二
  1. IT之家·人工智能52

    英特尔 High-NA EUV 晶圆处理量突破 100 万片,超越其他厂商总和

    英特尔宣布其 High-NA EUV 设备累计处理的 300mm 晶圆突破 100 万片,距首台设备完成组装不到两年半,处理量已超过行业其他用户总和。该技术已用于 Intel 18A 工艺并量产 Panther Lake 酷睿 Ultra 处理器,目前仍受 6×6 英寸光罩尺寸限制,英特尔正推动行业转向 6×12 英寸光罩以实现 26×33mm 完整视场。

  2. vLLM 官方博客63

    vLLM x AgentX:面向真实世界智能体推理服务的全栈优化

    vLLM 团队发布针对智能体负载优化的博客,基于 SemiAnalysis 的公开 AgentX 基准展示成果:DeepSeek V4 Pro 在 GB300 上达到 83K total tokens/GPU-秒。

    推荐理由:vLLM 团队结合 AgentX 实测数据讲解智能体负载的全栈优化方法与失败教训,读者可以借此理解并行策略如何跟随模型架构调整。

  3. Simon Willison 博客61

    kernel.org 运维者谈滥用爬虫:14 个 CPU 核心专为爬虫渲染 git 提交页面

    Simon Willison 转述 Konstantin Ryabitsev 的观察:git.kernel.org 为应付滥用爬虫的“背景辐射”,在 5 个地理分布节点上常年有 14 个 CPU 核心专门把 git 提交渲染成 HTML,CPU 开销超过包括 git clone 在内的所有合法访问。Willison 担忧这对提供大量可爬网页的 Datasette 也有影响。

  4. The Decoder:AI News77

    Anthropic 据报道签约高达 5170 亿美元算力协议,锁定至少 14.8 GW 算力

    据 The Information 报道,Anthropic 在十一个月内签署了价值高达 5170 亿美元的算力合同,自 2025 年 10 月以来锁定至少 14.8 GW 算力,并计划自建数据中心。

    推荐理由:两家公司在算力扩张上的表态与实际动作形成对照,读者可以借此审视巨额算力承诺与收入规模之间的差距。

  5. MiniMax (official)64

    NVIDIA Sol 团队与 MiniMax 合作发布 Sol-H3 端到端推理加速栈,在 8× NVIDIA B300 Blackwell 上生成 5 秒 1344×768 立体声视频仅需 1.653 秒,相比 50 步 Base H3 最高提速 15.54×,且在 1×、4×、8× B300 上均可扩展。

    引用Enze Xie@xieenze_jr

    🚀 Sol-H3: @MiniMax_AI H3 Video Generation Faster Than Playback 🤩 Five seconds of world. 1.653 seconds to infer. We’re releasing Sol-H3, our fastest end-to-end MiniMax-H3 inference stack yet. On one 8× NVIDIA B300 Blackwell system, it generates five seconds of 1344×768 video with stereo audio in 1.653 seconds. Across 1×, 4×, and 8× B300, Sol-H3 reaches up to a 15.54× speedup versus Base H3. Compared with 50-step Base H3 Dense on the same 8× B300 system, the four-step Sol-H3 profile delivers: • 5s: 18.250s → 1.653s (11.04×) • 10s: 50.660s → 3.732s (13.57×) • 15s: 99.513s → 6.612s (15.05×) Sol-H3 also scales across GPU counts: • 4× B300: 2.918s / 6.993s / 12.542s for 5s / 10s / 15s (12.11–15.54×) • 1× B300: 13.745s / 37.813s / 52.260s for 5s / 10s / 15s (9.45–14.29×) All figures are medians of three measured runs after one warmup at 1344×768 and 24 FPS with stereo audio. Base H3 uses 50 scheduler points (49 DiT forwards); Sol-H3 uses four DiT forwards, so this is a full-profile comparison—not an attention-only runtime change. Sol-H3 uses Dense attention on 1× B300 and SOL with INT8 QKV / FP8 output transport on 4× / 8×. Timing includes text encoding, DiT denoising, and video/audio VAE decoding; model loading, compilation warmup, and final MP4 encoding are excluded. Sol-H3 brings Sol-Engine × Sol-Attn into one full-stack runtime: • dynamic sparse attention with no retraining • fused norm, RoPE, MLP, and sparse-attention setup • fused INT8 QKV / FP8 output communication across 8 GPUs • parallel, batched VAE decoding • precomputed AdaLN caching Inside the stack: • sparse-attention setup: 1.206 → 0.285 ms (−76.4%) • VAE decode: 7.55 → 0.602 s • ~24 GB memory freed per GPU Any MiniMax-H3 few-step LoRA can plug into the same engine, and the code is deployment-friendly under Apache 2.0. For us, the bigger milestone is crossing from “fast generation” into “faster than playback.” That opens the path toward continuous 24 FPS generation and truly interactive video systems. We’re excited to partner with @reactorworld to release Sol-H3 and make it available as an API day-0. Try it now on Reactor: https://www.reactor.inc/sandbox?model=fast-h3 🔗 https://nvlabs.github.io/Sana/Sol-Engine/Sol-H3/ Amazing team effort—full credits in the blog. @shanasaimoe @lawrence_cjs @yitongli165665 @haopengl33 @HaochengXiUCB @songhan_mit

  6. OpenBMB56

    面壁智能 OpenBMB 宣布 MiniCPM5-2B 就绪,可通过 SGLang 部署,SGLang 提供首日(day-0)服务支持。SGLang 配方覆盖 NVIDIA RTX PRO 6000、RTX 5090、DGX Spark 和 Hopper 卡,在 5090 上启用 DSpark 后编码任务 bs=1 下单用户解码速度超过 250 tok/s。

    引用SGLang@sgl_project

    Congrats @OpenBMB on launching MiniCPM5-2B! SGLang has day-0 serving support, with recipes for @NVIDIAAI RTX PRO 6000, RTX 5090, DGX Spark, and Hopper cards. We are seeing over 250 tok/s/user decode at bs=1 on coding tasks with DSpark enabled on a 5090. It's an ultra-efficient 2B with incredible intelligence density. As of today it leads Artificial Analysis's Intelligence Index among open models under 4B, and scores above models more than 4x its size. This is another step toward general-purpose models running 24x7 on device, even on phones. Run MiniCPM5-2B with SGLang 👇

  7. Cohere Labs:官方研究博客68

    Cohere 详解 North Mini Code 的 megakernel 推理引擎,单 H100 上比 vLLM 快 1.25–1.41 倍

    Cohere 发布为 North Mini Code 构建的围绕 decode megakernel 的推理引擎,BF16 下单张 H100 端到端解码吞吐比 vLLM 快 1.25–1.41 倍,batch size 1 时达 292 tok/s(SoL 的 62%),代码已在 GitHub 开放。

    推荐理由:原文给出手写 decode megakernel 的完整实现细节与真实吞吐数据,读者可以了解如何把现有内核组装成单个持久内核并接入真实服务。

9月7日周一
  1. IT之家·人工智能55

    摩尔线程触及 20cm 跌停,股价创上市以来新低,市值跌破 2000 亿元

    摩尔线程 9 月 7 日早盘触及 20% 跌停,股价创上市以来新低,总市值跌破 2,000 亿元。同日其 2,577.45 万股首发网下配售限售股解禁流通,占总股本 5.48%。公司 2026 年上半年营收 17.36 亿元,同比增长 147.42%,归母净亏损 1,156.31 万元,同比收窄 95.73%;另于 8 月 9 日公告拟发行 H 股并申请在香港联交所主板挂牌上市。

9月6日周日
9月5日周六