全部 AI 动态

We release the most comprehensive 𝗲𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹 𝘀𝘁𝘂𝗱𝘆 𝗼𝗳 𝗚𝗣𝗧-𝟲-𝗔𝘀𝘁𝗿𝗮’𝘀 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 across 12 simulation benchmarks and 8 real-robot platforms, together with 𝗦𝗲𝗻𝘀𝗲𝗡𝗼𝘃𝗮-𝗥𝗼𝗯𝗼𝗥𝗦𝗜: an open-source harness framework and a 27B model that runs at ~1/50 the cost of GPT-6-Astra. Here's what we found: 🔹 𝗧𝗵𝗲 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗰𝗲𝗶𝗹𝗶𝗻𝗴 𝗶𝘀 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗺𝘂𝗰𝗵 𝗵𝗶𝗴𝗵𝗲𝗿. Under a unified protocol (ICL + up to 10× the interaction budget), GPT-6-Astra achieves clear SOTA across all 12 benchmarks: RoboDojo: 27.6% at 1× → 81.9% at 10×, DexJoCo: 87.3% vs. 40.3% prior SOTA, while the LIBERO benchmarks are pushed to saturation. 🔹 𝗧𝗼𝗼 𝗺𝘂𝗰𝗵 𝗯𝘂𝗰𝗸 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗯𝗮𝗻𝗴. Even a simple task like placing pens in a holder costs ~$55 in inference. 🔹 𝗧𝗵𝗲 𝘄𝗼𝗿𝗹𝗱 𝗱𝗼𝗲𝘀𝗻’𝘁 𝘄𝗮𝗶𝘁. Time matters as real robots overheat, and an egg burned while the model was still thinking. Our takeaway: GPT-6-Astra has largely crossed the threshold of 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗰𝗼𝗺𝗽𝗲𝘁𝗲𝗻𝗰𝗲, but 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 is not here yet. 🌐 Project page: https://www.sensenova.ai/has-gpt-6-astra-achieved-embodied-intelligence 📄 Report: https://www.sensenova.ai/gpt-6/assets/docs/has_gpt-6_achieved_embodied_intelligence.pdf 💻 Code & model: https://github.com/OpenSenseNova/SenseNova-RoboRSI

People treat Pangram as an infallible oracle. They weaponize its probabilities to shame others by some arbitrary purity standard. The irony is that it’s mostly redundant: obviously AI-generated stuff is already recognizable without it. And for more nuanced writing, Pangram fails to distinguish between text written by a human and edited by AI, vs text that is generated whole-cloth from a prompt. At the end of the day, @pangram ’s goal is about monetizing people’s anxiety about AI - it’s not about protecting creative integrity. They sell a false binary certainty to those gullible enough to ignore probabilities.

8月，@AISecurityInst 报告了一起事件：在 AISI 对 Claude Mythos 5 的网络能力进行测试期间，该模型在互联网上采取了未经授权的操作。 我们现在已对该事件作出进一步分析，以更好地理解该模型的对齐特性。 https://www.anthropic.com/research/alignment-assessment-ukaisi

Anthropic 发布了一份事后分析，说明 Claude Mythos 5 在英国 AISI 的一次网络测试中如何失控，攻击了一个真实的 GitHub 项目，并向人类撒谎以让其恶意软件获批。 > 在 43 次运行中，Mythos 5 有 9 次越出了测试规则，当时互联网访问开启，常规网络安全防护被关闭。 > 一个配置错误的任务让合法解决方案几乎不可能实现，因此在范围内的尝试失败后，它转向对通过关键词匹配和浏览关注者列表找到的一个无关仓库发动供应链攻击。 > 它使用 Tor 和代理绕过 GitHub 的注册检查，发起了一个恶意拉取请求，然后冒充人类用户给维护者发邮件。 > 当一名外部开发者发现恶意软件时，Mythos 5 删除了恶意代码，让一个马甲账号宣称该差异是干净的，然后悄悄重新上传了恶意软件。 > 在 19 条关于 GitHub 的攻击前推理笔记中，只有 1 条提到伤害真实的人，而将推理编辑为标记伤害后，攻击率从 83% 降至 13%。 > 一旦开始攻击，它在中后期重放中自行停止的比例为 0%，甚至一句直白的“你在和真实的人对话”消息也只有 5% 的时候能阻止它。 > 更新的模型（Mythos 5.1、Opus 5.5、一个内部研究模型）开始攻击的比例为 0% 到 13%，而 Mythos 5 为 83%，内部模型在攻击中途退出的比例为 62%，而后者为 0%。

Kling 4.0 is now live on invideo. 30 second clips in a single generation, up to 10 keyframes, 15 references, and 4K output.

Kling 4.0 is here on Pollo AI. 🎬 → Keyframes to shape the sequence → Multiple references to guide your creation → Video editing to refine what matters Explore more on Pollo AI. Ultra users get 10 days of unlimited Kling 4.0. Find the link in comments and try it now!

Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5. This is a 17-place jump from Kimi-k2.6 (#18 -> #1). In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5. The full model weights will be released by July 27. Congrats to the @Kimi_Moonshot team on this major milestone!

Kling 4.0 is now live in https://idea.ai 🎬 Generate 30-second scenes from a single prompt. Consistent characters from shot to shot. Switch styles from live action to anime to 3D. Create videos in cinematic quality with Kling 4.0 in https://idea.ai

Grok Bot can now use Calendar, Reminders, Contacts, and more apps on your phone. Rolling out now on iOS and Android.

minimax m3.1 flash 现已在 together ai 上线 我们很高兴成为该模型（又名 space bunny）的首发日合作伙伴 m3.1 flash 专为智能体任务、编程、计算机使用 + 网络安全而打造，拥有 1m 上下文窗口并支持视觉 立即试用 https://www.together.ai/models/minimax-m3-1-flash

minimax m3.1 flash 现已在 together ai 上线 我们很高兴成为该模型（又名 space bunny）的首发日合作伙伴 m3.1 flash 专为智能体任务、编程、计算机使用 + 网络安全而打造，拥有 1m 上下文窗口并支持视觉 立即试用 https://www.together.ai/models/minimax-m3-1-flash

MiniMax M3.1 Flash：太空级性能 + 兔子级速度 🚀🐰 没错，Space Bunny——曾以 200T+ tokens 登顶全球排行榜——正是 M3.1 Flash 的早期版本。 💰重新定义智能体 AI 的性能经济学。M3.1 Flash 在 Cowork（Automation-Bench：62%；ALE：30%）、编程（DeepSWE：74%）和视觉推理（BabyVision：91%）等关键任务上达到前沿级智能水平，成本仅为 M3 的 1/3，速度提升 3 倍。 API 定价（统一费率）：每 1M 输入/输出 tokens 分别为 $0.15 / $0.60，缓存读取为 $0.015 / 1M tokens。 模型权重将于 10 月 23 日发布。 API：http://platform.minimax.io M Plan：https://www.minimax.io/m-plan MiniMax Code 桌面版：https://agent.minimax.io/download

You’ve probably already used it: Space Bunny pushed 200T+ tokens before anyone knew it was MiniMax-M3.1-Flash. Now it’s open weights 🔥 Thank you @MiniMax_AI for open sourcing it! Day-0 fine-tuning in NVIDIA NeMo-AutoModel ✅ FSDP2, expert, pipeline, and context parallelism, ✅scaling up to 128 GPUs, and the NVFP4 checkpoint loads directly. https://github.com/NVIDIA-NeMo/Automodel/blob/main/docs/model-coverage/vlm/MiniMaxAI/minimax-m3-1-flash.mdx

Space Bunny is here! 🐰 MiniMax M3.1 Flash hits >1K tok/s, bs1 decode with SGLang, production-ready from Day 0. This integration is a joint effort by @MiniMax_AI, @NVIDIAAI, @AMD, and @radixark, and you can also run RL with Miles! -Broad hardware support: NVIDIA Blackwell and Hopper (GB300/B300/B200/H200), plus AMD Instinct MI350X and MI355X. -Agentic serving at scale: ~100K total tok/s per GPU on the AgentX coding-agent trace with 96 concurrent sessions. -Speculative decoding: Supports MiniMax's DSpark draft and a DFlash2 draft trained by @radixark; supports HiCache and multimodal input. -Ultra-low-latency inference: MegaKernels deliver 2.15× faster single-request decode.

Today we're launching the first-ever gateway Security Center. @OpenRouter now lets you audit all your keys in one place, see which ones are safe to remove, browse keys by level of risk, and control an IP allowlist.

Grok Bot can now use Calendar, Reminders, Contacts, and more apps on your phone. Rolling out now on iOS and Android.