Falcon-TST 2.0: Quantile Time Series Forecasting in Python
Forecast with Falcon-TST 2.0: open source, tops GIFT-Eval at MASE 0.666. pip install falcon-tst, run quantile_predict, get 21 quantiles.
Falcon-TST 2.0 上手：Python 分位数时序预测
蚂蚁国际开源的时序基础模型 Falcon-TST 2.0 在 GIFT-Eval 基准拿下第一（MASE 0.666）。一行 pip install 即可用，API 直接输出 21 个分位数，适合外汇、供应链等场景做风险预测。
OpenAI's Jalapeño Chip Beats NVIDIA: First Benchmarks
OpenAI first chip Jalapeno beats Nvidia GB300: 1.5-1.9x work per watt, 3.6x lower latency, GPT-6 co-designed the silicon in 9 months.
OpenAI 首颗芯片 Jalapeño 实测：推理碾压 GB300，CUDA 护城河告急
OpenAI 首颗自研芯片 Jalapeño 首批实测：每瓦工作量高 1.5–1.9 倍、延迟低 1.7–3.6 倍，全面反超英伟达 GB200/GB300；设计到流片仅 9 个月，GPT-6 参与设计芯片。CUDA 护城河之争，从口号变成可测量数据。
AI Daily Briefing – 2026-08-26: Embodied AI Goes Co-Built
Daily digest of the most consequential AI developments, written for a global reader. 🔥 Top 3 Highlights 1. Phanthy and UBTech rally a dozen-plus players to co-build PhanthyMotus What happened: Phanthy (范式) held the launch of its PhanthyMotus ecosystem co-build program, joined by UBTech and more than ten embodied-AI leaders — moving its general-purpose embodied Agent foundation from "open source
AI 早报-2026年8月26日
每日精选全球 AI 资讯，快速掌握行业趋势 🔥 今日亮点（Top 3） 1. 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 核心洞察：新产品密集推出，竞争加剧 来源：https://www.qbitai.com/2026/08/479314.html 2. 开源国产8B模型，比肩闭源Image 2了！ 核心洞察：国产模型快速崛起 来源：https://www.qbitai.com/2026/08/479192.html 3. 半年3轮10亿，他们都投了这家已经把机器人卖到500个家庭的公司 核心洞
AI Wellbeing Evaluations: Anthropic Funds Open Benchmarks
Anthropic funds open-source wellbeing evals with a $5M grant — from capability to what AI does to people. Applications due September 21.
Anthropic 500 万美元押注福祉评测：AI 基准从「能力」转向「对人的影响」
Anthropic 启动 500 万美元资助计划，支持独立团队开源建设 AI 福祉评测基准：衡量从「模型能做什么」转向「模型对使用者做了什么」。多轮对话、临床专家、过度顺从与过度拒绝对称测试……福祉正在变成工业级规格。申请 9 月 21 日截止。
AI Token Cost, Explained by NVIDIA's Five-Layer Cake
The five-layer cake: chips take 40-55% of AI token cost — and a double clock explains why price cuts and compute scarcity coexist.
一个 Token 到底多少钱？黄仁勋「五层蛋糕」拆穿 AI 成本真相
黄仁勋「五层蛋糕」拆穿 AI 成本：芯片约 40%-55% 是大头，能源/基础设施/模型/应用各占 10%-20%/15%-20%/10%-15%/5%-10%；「双时钟」定律解释价格战打到骨折、算力却始终紧缺。附 5 条落地建议。
AI for Science Moves from Tasks to Projects: AutoProject
AutoProject moves AI4S from tasks to projects: planning, self-repair, and evidence-graph verification — tool-level to system-level science.
AI4S 进入「项目时代」：AutoProject 把 AI 科研从 Task 推向 Project
中科紫东太初升级 ScienceClaw 推出 AutoProject 引擎：Project2Task 规划、TaskExecutor 长程执行、EviGraph 证据验证，AI 科研从 Task 走向 Project，能力从工具级走向系统级。