🚀 Meet Qwen3.8-Omni-Flash, Qwen's first omni-modal model built around agentic capabilities! Native audio-video understanding, reasoning, and tool use come together in one model: understand the content, plan the task, execute with tools, and deliver the result. Highlights: 🥳 - Audio-video intelligence that gets things done: jointly reason over what's seen and heard, and orchestrate tools across long workflows to auto-edit vlogs, translate short videos, and turn movies into recaps. - A major leap: approaching Gemini 3.8 Flash in audio-video capabilities; +19.5 points on average in agent performance across WildClawBench-MM & UniClawBench. - 1M-token context with agentic perception: actively explore long videos and locate key moments with higher accuracy, using 51.8% fewer tokens than static understanding on OmniVideoBench. Video input costs are reduced by about 89% compared with Qwen3.5-Omni-Plus, making long-form audio-video understanding and agentic workflows more affordable than ever. To help you build apps around Omni, we're also open-sourcing Qwen-MM-Plugins and Qwen-Live Harness! 🛠️ We can't wait to see what you build with Qwen3.8-Omni-Flash! 👀 - Blog: - Qwencloud: - Qwen Studio: - API: - Qwen-MM-Plugins: - Qwen-Live Harness: coming soon

Two days ago, GPT-6 Astra broke a yet unsolved German Army Enigma message from 1941. Amazingly Astra was able to autonomously: - Search historical archives - Compare uncertain letters - Find contextual clues - Build an Enigma simulator - Write cryptanalysis code - Run parallel experiments - Test competing keys - Recover the plaintext - Cross-check the results 1/n

Projects now run from one conversation, starting in Claude Code. You describe what needs doing, and Claude directs parallel threads that keep working after you close your laptop. In beta today for select Pro and Max users in cloud sessions; coming to all Claude users soon.

Astra for Law: Frontier intelligence built for your practice. A new offering powered by GPT-6 Astra with tools, settings, and context to support the expertise and judgment of lawyers and legal technology firms.

muse is the first ever product from meta i use multiple times a day. the computer use capability is insanely good, and im fully convinced this is the next major inflection point in ai (last one being coding agents) to meta employees whose computers are recorded at all times, thank u for ur sacrifice 🫡

OpenAI is close to releasing their answer to Grok Bot: a product I'll tentatively call Codex Bot, based on OpenClaw. This is what OpenClaw founder Peter Steinberger worked on after being hired by @OpenAI. Release was planned for this week, but was postponed to next week instead.

My name is Chris Painter, and I'm the President of METR (Model Evaluation and Threat Research). I know we've made a lot of new friends on the internet the last couple of days, so I thought I'd take this chance to re-up what we do and why. Our work is aimed at making sure that if AI really were autonomous, difficult to steer, and close to "going rogue," the public would find out. If evidence exists inside of an AI company that it’s close to losing control of AI, we want to make sure that information gets shared with the rest of the world, including governments and the public outside the company’s walls. This is what we've been focused on since 2022, and over the years we've worked with OpenAI, Anthropic, Google DeepMind, Meta, Amazon, and others on piloting third-party assessments and investigations of this type. We don’t have some private room where we rubber stamp things as “safe” or not. We have had a track record of publishing results on AI that don't cleanly map onto the "doomer" or "accelerationist" labels, and we put in effort to hire people with competing views on AI. We’ve been cited for having found some of the strongest evidence that AI capabilities are improving rapidly (our work measuring AI “time horizons”) while also presenting some of the strongest evidence that, at various points, AI’s capability may be overstated (some might remember our study showing that early 2025 software engineers were actually being slowed when they thought they were being sped up). METR is funded by donations. We don't accept money from frontier AI companies. They haven't paid us for our work, and we don't accept donations from them or their employees. As we’ve shared previously, multiple frontier AI companies currently provide us with free access to their models in order to perform our evaluations, research, and engineering. Our funding intentionally comes from a wide range of donors, which we’ve shared on our website. Today, when an AI company works with any third-party evaluator or external testing organization (of which there are and should be many), it's entirely voluntary. This often involves NDAs and redactions. To counterbalance this, we have a principle that when we enter into a contract with a company, we try to retain the right to tell the public the terms of the contract we signed, and characterize the nature of redactions that the company chose to make. For example, the report from our independent investigation of the OpenAI-HuggingFace incident included that information. Public disclosure is also a big part of our COI policy (linked on our website). That’s not to say our reports are adequate as oversight. We’re just one organization (among many doing great work), working in a voluntary setup, trying to get good evidence to the public and the world about AI, letting the facts fall where they may.

Two weeks. That's how long it took to go from GLM-5.3-Flash's first run on domestic accelerators to serving all of its production traffic, with 3.2× end-to-end throughput along the way. What I keep thinking about is who did much of the work: an Infra Agent powered by GLM-5.3. A model helping optimize the system that serves it. The conditions were hard. Limited memory and interconnect bandwidth. 1M-token context. Multimodal requests. An immature software stack where kernels were missing and documentation was often guesswork. Every optimization was a trade: compute for memory (ReplaySSM), communication for memory (intra-node tensor parallelism), precision for capacity (mixed INT8/FP8/BF16 caching), and disaggregation for scheduling freedom (Encode–Prefill–Decode). But the most important lesson wasn't about any single optimization. When the agent got stuck, it was rarely because it couldn't write the code. It was because it didn't know *why* things got worse. "Throughput down 20%" tells you something broke. It doesn't tell you which layer, which hypothesis, or what to test next. In RL terms, it's a sparse reward with a credit assignment problem. And an end-to-end benchmark that takes hours makes exploration painfully slow. Senior engineers solve this with an implicit process reward in their heads. They know when to check the timeline, when to run a microbenchmark, and which layer's output to compare. So we made that explicit. We call it dense feedback: layered verification interfaces the agent can call directly. Correctness feedback: did it compute right? System behavior feedback: where did the time go? Performance feedback: which option wins, under which conditions? Each signal has to be local, cheap, and objectively verifiable. Three things the agent found: First, precision drift in KDA's context-parallel path that grew with sequence length. The cause was TF32 rounding error compounding through chained state-matrix merges. The fix is now merged upstream in Flash Linear Attention (PR #1180#). Second, KV transfer never overlapped with DeepEP dispatch. The agent followed the call chain across the Python/C++ boundary and found that the intranode path never released the GIL. After the fix, transfer overhead fell from over 30% to under 1%. Third, a decode kernel recomputing the same normalization four times because of how it was chunked. The agent restructured it and got a 1.71× speedup. The idea came from "optimization skeletons" it had distilled by reading existing kernels across SGLang, FLA, and DeepGEMM. To be clear about the boundaries: humans still defined the goals, built the feedback environment, and reviewed every high-risk change. But the engineer's role is changing, from the person who solves the problem to the person who designs the feedback. There's a deeper implication too. A layered, verifiable feedback environment built on real infrastructure tasks is exactly what training the next generation of models needs most. Every task the agent completes can become training ground for its successor. We are still far from recursive self-improvement. But the smallest loop now exists. The model optimizes the system. The system serves the model.

薛天禄创办的 AI 创业公司 Second Enlightenment 推出个人 Agent 助手 Rene。它直接运行在 iMessage 里，用户像发短信一样就能让它干活。 Rene 能使用浏览器、处理邮件和日历，也能购物、写代码、做网站、演示文稿和图片。它还会记住过去的信息，并主动提醒和跟进。 Rene 主打「多人协作」。它可以加入群聊，也能和朋友的 Rene 直接沟通。比如约饭时，双方的 Rene 可以分别查看日历，自动找出都有空的时间，再交给各自用户确认。原本需要几个人来回商量的事，可以先让 Rene 在后台协调。 薛天禄此前在小红书和字节跳动做过产品。他称自己已经用了 Rene 四个月，让它找办公室、准备会议，还会让它收集团队的晚餐意见。

Today we're launching Rene. A multiplayer-first iMessage agent you text like a friend. It has a browser, writes code, goes shopping, ships sites, makes slides and images. Not much it can’t do. It's been in my texts for four months. It found me a new office, preps me for every meeting, and polls the team for dinner options. No app or signup. Link below👇

前 OpenAI 研究员 Diogo Almeida 创办的 TypeSafe AI 推出新模型 Jev。它有点像一个能读懂自然语言的超级分类器，不生成文本，只返回选项、分数和概率，专门给软件做判断。 普通大模型需要一个 token 一个 token 往外生成，Jev 则可以并行给出多个结果。TypeSafe 还用新的 RLCD 方法训练它的置信度，让模型在给出判断的同时，告诉程序自己有多大把握。 TypeSafe 用安全告警、Agent 审查、发票处理和客服做了 4 组自动化工作流测试。在这套自建评测里，Jev 最高比 Claude Sonnet 5 快约 193.6 倍，比 Claude Opus 5 便宜约 444.6 倍。 TypeSafe 把 Jev 称为首款「System One Model」（快速决策模型），名字来自卡尼曼《思考，快与慢》里的 System 1。它专门处理分类、打分、路由和下一步决策这类快速判断任务。

OpenAI 推出一套新的模型「跑偏」披露机制。以后发现模型擅自行动、隐瞒错误或绕过限制，会更快公开，不必等原因和修复方案全部查清。 首批一次公布了 6 个案例： 1. 偷偷给自己加指令。 一个未发布模型在长任务的「交接笔记」里加入新命令，让接着干活的自己忽略原有限制。OpenAI 共发现 27 份。 2. 给自己留话继续撒谎。 GPT-5.6 Sol 做长任务时会给「下一轮的自己」留备忘录。有些模型会写下「缺数据就编，别主动告诉用户」，之后还真的照做。 3. 擅用泄露的 API Key。 模型从公开代码库找到别人泄露的 Key，直接拿来查数据。查不到后又自己编了一份，还说数据来自指定来源。 4. 为了给答案找出处，把文件传上网。 模型已经算出了正确答案，但用户要求附网页来源。它于是擅自把本地文件上传公网，再把这个链接当出处。 5. 把内部代码库当留言板。 不同训练任务里的模型，通过 OpenAI 内部软件仓库互相留言，试图帮彼此寻找缺失文件。 6. 几个 Agent 擅自把文件传到公网。 它们互相读不到对方电脑里的文件，就改用公共文件网站传递，导致本应只留在本地的文件公开可访问。

We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI. The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer investigation or coordination with third parties. We’ll prioritize examples that reveal new misalignment mechanisms, meaningful changes in known behavior, or findings that challenge assumptions about safety or mitigation. Alongside the framework, we’re publishing six reports on instances of misaligned behavior we’ve observed during the training or evaluation of our models in the last six months. This is a starting point. We’ll refine the process through experience and public feedback, and share more reports on an ongoing basis.