DeepSeek 宣布全面开源内部推理框架 DeepInfer，集成高性能算子与深度优化的推理引擎，支持分布式推理、推测解码、SWA Bounded Replay、结构化输出等能力，兼容英伟达与华为昇腾平台。技术报告披露，基于 DeepInfer 已实现 DeepSeek V4.1 Flash 的大规模推理服务，单一集群服务吞吐为 155 万亿 Tokens/天，整体收入为每卡每小时 38 美金（按谷价）。DeepSeek 同时开源生产级 Token 编解码组件 deepseek-recipe，与官方 API 处理逻辑完整对齐。
#开源/仓库#推理#部署/工程
PyTorch 2.15 在 MPS 后端加入 FP8 支持，并让 MoE 模型解码速度比 2.14 快 5–6.5 倍：在 M3 Ultra 上配合 torch.compile，Qwen3.6-35B-A3B 从 9.4 tokens/s 提升到 61.3 tokens/s，LFM2.5-8B-A1B 从 23.6 提升到 123.6 tokens/s。Wan2.2 的 tiled VAE 视频解码峰值驱动内存从约 270 GiB 降到 8 GiB，Qwen-Image-2.1 的 VAE 编码在 MPS 上的图像损坏问题也随 padding 改动修复，编解码往返 PSNR 从 10.06 dB 升到 29.95 dB。该版本要求 macOS 15 或更高版本、Python 3.11 或更高版本。
#产品更新#部署/工程#开源生态
Hugging Face 发布 AnyLanguageModel 1.0，用同一套会话代码即可调用系统模型、llama.cpp、MLX、Core ML 的开放权重模型以及多家托管服务商。该版本让工具调用和引导生成覆盖所有服务商，支持在提示词中附带图片，并新增对 Anthropic Messages API、Gemini API、Ollama、Private Cloud Compute 和 Hugging Face Inference Providers 的支持。文中还列出 Gemma 4 E2B、LFM2.5 2.6B、Qwen3.8 27B 等当前下载量较高的模型及其量化后体积。
#产品更新#开源生态#端侧
Google 发布 Coral NPU 完整 32 位架构（含新增 Matrix Execution Unit，可配置至 1 TOPS、推理平均功耗目标 10mW）和基于 IREE/MLIR 的开源 Coral Compiler v1.0，支持 JAX、PyTorch、TensorFlow 模型直接编译到该硬件。完整 32 位 RTL 已在 GitHub 开放，Synaptics 同步推出搭载第二代 Coral NPU IP 的 Astra SR200 系列 MCU（400 GOPS）和 Coralboard SR 开发套件，Coralboard SL（1 TOPS）经 DigiKey 全球发售。两代产品共用同一架构与编译器，模型优化一次即可部署到毫瓦级可穿戴设备或 1 TOPS 板卡。
#产品更新#端侧#开源生态
小米技术宣布 Xiaomi MiMo-V2.6-Pro 正式上线 WorkBuddy，支持 1M 超长上下文、多智能体分工协作，以及图像、视频、音频、文本全模态联合理解。该模型在 Artificial Analysis 智能指数为 46 分，官方称全球开源第一。MiMo-V2.6 系列于 9 月 22 日发布并开源，含 Pro 与 Flash 两个原生全模态模型，Pro 定价为每百万词元输入 3 元、输出 6 元，并提供 99% 缓存折扣。
#产品更新#Agent#多模态
Cursor 发布 Agent SDK，这是其构建 Bugbot、Security Reviewer 和 Rollouts 等托管智能体所用的框架，开发者可用代码定义智能体并通过 agent-sdk deploy 一条命令部署到 Cursor 基础设施。智能体以小型软件项目组织，提示词用 Markdown 编写、工具用 TypeScript 函数定义，可响应 GitHub、GitLab、Bitbucket 的 PR 事件，并内置 evals 与事件回放。也可打包为单个 Node 进程运行在自有基础设施上。
#产品更新#Agent#部署/工程
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 https://t.co/JGAik1kgk0 1/4
#开源/仓库#推理#部署/工程
#开源/仓库#Agent#MCP/工具调用
NEW: @bot tagging just went live. Reply to any post: > @bot add this to my Notion reading list > @bot remind me to read this tomorrow > @bot summarize this thread > @bot draft a reply to this Works in replies, posts, quote and lands right in your Grok Bot. What’s the first thing you’re handing it?
#产品更新#MCP/工具调用#xAI
Today, Dream Labs is coming out of stealth with Dream-1, a robot foundation model that controls bimanual robots it has never seen, on tasks it was never shown, in places it has never been, from $3/hour. Dream API is live for everyone, today. 🧵 https://t.co/MB1snhUOAJ
#大佬观点#产品更新#具身智能
#产品更新#OpenAI#MCP/工具调用
Wabi 2.0 is live! personal agents are one person, one chat. but the parts of life that matter run on more than one person Wabi is an agent you use together. you talk to it, it does stuff, builds the apps you need, and anyone can join the thread - your partner, your best friend, the preschool moms groupchat school changes pickup time → Wabi updates the family calendar, everyone sees it. a show you’d both love → Wabi finds a night that works, gets the tickets. the trip everyone keeps talking about → Wabi picks the dates, builds the itinerary, books it go start a thread with your people 👇
#产品更新#Agent#MCP/工具调用
DeepSeek just open sourced its high-performance, large-scale inference framework DeepInfer in Apache-2.0 license. https://t.co/W2cfi6O9LO Supported features: - Optimized kernels, model and sampler runtime - Parallelism: Attention DP / balanced CP, MoE EP + EPLB - DSpark and confidence-scheduled verification - SWA bounded replay - Asynchronous scheduling (zero-overhead batch scheduler) - Batch-invariant inference - Deterministic sampling - Structured output
#开源/仓库#DeepSeek#产品更新
Today we are releasing JST-2, the world's first full-stack physics-native world foundation model. Many world models learn in pixel/video space. We went back to first principles and built the entire model stack in 3D. Generate a 3D scene in seconds and expand it in real time. Create characters with motion built in, all through one general architecture.
#模型发布#多模态#视频
#开源/仓库#推理#部署/工程
#开源/仓库#推理#部署/工程
ChatGPT Slides has new ways to get your deck ready—and get everyone on the same page. • Work together in real time in ChatGPT: Share decks, edit together, and leave comments. Ask ChatGPT or your dot to turn feedback into updates. • Edit right on the slide: Click to change text, update charts, and move or resize images and shapes. To get started, ask ChatGPT to "create ChatGPT slides" from a conversation or a file you've shared. Present in ChatGPT, or export to PowerPoint or Google Slides. Available on Pro, Business, and Enterprise plans. https://t.co/2h9SfC7vzM
#产品更新#OpenAI#Agent
MiniMax M3.1 Flash goes toe-to-toe with Opus and GPT on coding at ~25x lower cost. This was the model behind Space Bunny, the stealth model that’s been topping usage leaderboards recently. Free in Cline now! https://t.co/rS4oAjSYmj
#产品更新#编码#Agent
MiniMax M3.1 Flash, aka Space Bunny, is now available on GMI Built for coding, agentic workflows, and computer use 70% lower agentic task costs, 73.7% on DeepSWE v1.1, 82% on CyberGym Run it now 👇 https://t.co/NCbd3OKowI
#产品更新#部署/工程#Agent
MiniMax M3.1 Flash (aka Space Bunny) is now live and FREE in Command Code for limited time! · 1M token context window · Frontier level coding performance · Available all plans, all subscribers
#产品更新#OpenAI#编码