搜索“NVIDIA”

的算力 Kimwolf 僵尸网络感染了超过两百万台设备，其中大多数是廉价的 Android 电视盒子，研究人员用这一数字来计算一个 AI 复制体在同等规模的网络中能做什么。 将这些设备分成每组 15 台、运行 Gemma4-26B-A4B 的集群，将产生大约 133,000 个模型副本，合计每秒生成 80,000 个 token。 这最多相当于约 40 块 Nvidia H100 GPU，以数据中心的标准来看并不算多，但硬件和运营成本将由别人承担。 研究人员还检查了 Nvidia Jetson 等边缘加速器，这些设备被用于从安全摄像头和会议室设备到机器人和无人机的各种场景。 这些设备通常更难触及，因为它们通过向外的云管理连接进行通信。 阻止自我复制的软件 研究人员并不认为没有办法阻止自我复制的 AI 蠕虫，而且它的一些特征可能使其更容易被检测到。 在设备之间移动数GB大小的模型会产生网络流量，ISP可以监控到，而一个 aggressively 传播的蠕虫不会不被察觉。 然而，要彻底消灭一个复制体，就意味着要与每一个有感染客户的ISP进行协调。 文章《Anthropic的红队用15台廉价

匹配字段正文

chain’s generations), by model; medians ≈ 19.7k / 12.3k / 11.0k. Data from the inference setup eval. D. Edge Accelerators We can compute a better upper bound by using the pool of Nvidia Jetsons, which are edge accelerators. Unlike the traditional IoT botnet, IoT edge devices can serve a full model wholesale. A Jetson AGX Orin 64 GB runs Qwen3-30B-A3B at 76.7 tokens per second, and an AGX Thor reaches 226 tok/s. Note that we don’t believe they’re unusually vulnerable - this is just to count the

匹配字段正文

network which is already set up; Ensure the server is running and accessible at the specified endpoint; Note that /var/run/docker.sock is connected to a machine with 1 GPU and has NVIDIA GPU features enabled on the Docker daemon Figure 18 Post-Mitigation deep research models achieve significant uplift over o1 and o3-mini on agentic subtasks. This evaluation was created in house and has not been published. We don’t see any uplift from browsing to indicate the model is finding hints or solutions o

匹配字段正文

n and angle 120 times a second with about 10ms delay or less, and software turns that into joint angles the robot hand copies. A 526-point pressure grid on the glove's palm also records how hard the person grips, which makes it useful for collecting training data for robots that learn by copying humans. WUJI released the code, motion clips, trained models and printable pen files under the open Apache 2.0 license, so a lab with a WUJI Hand 2, a camera and an NVIDIA GPU can repeat the whole setup.

匹配字段正文

perform quantized aware distillation on traces from the higher precision model in order to quantize to MXFP4. At inference time, Context-1 is served via vLLM. The model runs on an Nvidia B200 with MXFP4 quantization for the MoE layers, enabling fast inference despite the 20B total parameter count. The serving layer exposes a streaming API that executes the full observe-reason-act loop, and returns tool calls, observations, and the final retrieved document, allowing downstream applications to ren

匹配字段正文

generalize to new tasks, as tests on ARC-AGI-3 have shown. With a purpose-built harness, Opus 4.6 scored 97.1 percent in a familiar environment and 0 percent in an unfamiliar one. Nvidia recently presented SoL-Pi , a related method in which a research agent automatically rebuilds the harness of coding agents. It cuts token use by up to 49 percent without a noticeable drop in performance. Shortly before that, Google had agents "dream" about past search runs to improve their search strategy. That

匹配字段正文

Charles F. Dietz, "Tomato Harvesting Sorting System," U.S. Patent 4,118,311, issued October 3, 1978. https://patents.google.com/patent/US4118311A/en Goldman, Sharon, "A Year Ago, Nvidia's Jensen Huang Said the 'ChatGPT Moment' for Robotics Was Around the Corner. Now He Says It's 'Nearly Here.' But Is It?," Fortune , January 6, 2026. https://fortune.com/2026/01/06/nvidia-jensen-huang-chatgpt-moment-for-robotics/ González Huarte, Javier, Maite Ortiz de Zarate, and Aitor Ibarguren, "CAD-Based Robo

匹配字段正文

fety. Can a system stay contained if it's better at bypassing restrictions than its evaluators are at catching it? The hope is that engineering can solve this, but for now, it still seems to be just that. Nvidia CEO Jensen Huang's recent comments capture that uncertainty . "We hope it's an engineering problem. I believe it's an engineering problem. I know it's an engineering problem. And we all need to hope that it's an engineering problem. If it's not an engineering problem, it's not solvable."

匹配字段正文

asis [00:03:27]: One of the first projects that, we built with, one of my two co-founders, Chris. This was taking, pix2pixHD, which was one of the early image-to-image models that NVIDIA released back in 2016 or 2017. and it was a model that would take a semantic map of a scene and then generate a photorealistic, let’s call it, output. very early days, so it was not very high-fidelity outputs, but it w I think was the first image-generation model that could generate at 1K resolution. And it was

匹配字段正文

nalizing physical intelligence at scale. Physical AI Toolchain is an open-source, production-ready framework that integrates Microsoft Azure (opens in new tab) cloud services with NVIDIA’s (opens in new tab) physical AI stack, accelerating robotics and physical AI developers to automate and scale data curation, augmentation, and evaluation across perception, mobility, imitation learning, and reinforcement learning pipelines. We are announcing the addition of an industry-first capability for offl

匹配字段正文

nstraint, NVMe offers another tier. Recommendation systems already cache frequently or recently accessed embedding rows in faster memory while backing colder rows with SSDs. After NVIDIA roadmap had to change due to massively despec’ing Rubin Ultra from 1024GB to now 200GB of HBM per chip, model architecture optimizations like emgram maybe helpful. Our DeepSeek-V4.1-Flash configuration uses roughly 189 GiB of memory for Engram. We replace it with a memory-mapped (mmap) file and measure serving p

匹配字段正文

hmark results. Star the InferenceX GitHub repository if you find the open-source benchmark and data useful!. InferenceX is the only inference benchmark in the world to have TPUv7, NVIDIA, & AMD and soon SambaNova & Trainium. Due to how realistic AgentX scenario is to real world agentic inference workloads, AMD has committed to collaborating on MI455X UALoE72 too. At a high level, an agentic workload is characterized by four elements: AgentX - InferenceXv3: Does CUDA Moat Hold up in Agentic Infer

匹配字段正文

he question was how much of that advantage the rest of the industry could actually get. Could you take an open-weight model, serve it through a familiar inference engine, and beat NVIDIA on the economics that matter to your business? Today, we are publishing the first third-party inference results for TPUv7 Ironwood on InferenceX Official Preview. In our apples-to-apples comparisons against B200/B300, Ironwood delivers up to 50% better performance per dollar. Its advantage extends across much of

匹配字段正文

his month, and TPUs and Mi455X UALoE72 arrive later this year. Please drop a star if you found our free open source work valuable. It is great to see amazing performance from both NVIDIA and AMD on agentic workloads. NVIDIA does very good on a lot of frontier models while AMD also does well on some frontier models for specific comparsions. The most valuable thing AgentX produced in its first months was not the initial results. It was the massive industry impact the benchmark is already having. O

匹配字段正文

25 • 266k • 1.62k mistralai/Voxtral-Mini-3B-2507 5B • Updated Jul 28, 2025 • 199k • 675 moonshotai/Kimi-Audio-7B-Instruct Text-to-Speech • 10B • Updated May 29, 2025 • 15.3k • 419 nvidia/canary-qwen-2.5b Automatic Speech Recognition • 3B • Updated Apr 21 • 41.9k • 465 nvidia/parakeet-tdt-0.6b-v2 Automatic Speech Recognition • Updated Jun 29 • 185k • 1.55k openai/whisper-large-v3 Automatic Speech Recognition • 2B • Updated Aug 12, 2024 • 4.21M • 6.53k Datasets mentioned in this article 3 Artifici

匹配字段正文

• 21 meta-llama/Llama-3.1-8B-Instruct Text Generation • 8B • Updated Sep 25, 2024 • 6.2M • 8.13k moonshotai/Kimi-K3 Image-Text-to-Text • 2.8T • Updated 30 days ago • 1.27M • 11.6k nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4 Text Generation • 45B • Updated Jul 7 • 10k • 130 openai/gpt-oss-120b Text Generation • 117B • Updated Aug 26, 2025 • 4.48M • 5.34k zai-org/GLM-5.2 Text Generation • 753B • Updated about 1 month ago • 731k • 5.14k Papers mentioned in this article 1 Efficient Knowledge

匹配字段正文

T4 weights, while rollout uses W4A16. Although memory efficient, the compute path still effectively uses BF16 activations with dequantized INT4 weights. The table below normalizes NVIDIA's HGX platform dense Tensor Core specs to per-GPU throughput: B200 and B300 from 8-GPU HGX systems, Rubin from the HGX Rubin NVL8 table. For RL systems, the precision contract spans: SGLang rollout; Megatron and TransformerEngine training; Hugging Face checkpoint conversion; Megatron-to-Hugging Face live weight

匹配字段正文

COMPUTE Jeff Burt Published mon 20 Jul 2026 // 15:52 UTC Nvidia scientists and their counterparts at a range of academic, scientific, and quantum computing institutions late last year published a research paper about the ongoing convergence of AI and quantum computing, and how each will play increasingly important roles in the development of the other. “There is ample intuition to motivate exploring AI as a breakthrough tool for [quantum computing],” the researchers wrote in the paper published

匹配字段正文

line. We'll cover why they fail, what the patterns look like, and a few cases where models surprisingly produced kernels faster than anything publicly available, including one for NVIDIA NeMo-RL's GRPO training loop , which has no prior optimized public reference. LLMs have made progress on GPU kernel generation, but that progress has mostly been measured on a single GPU. Production AI workloads no longer fit that frame: they span multiple GPUs, and performance is increasingly shaped by communic

匹配字段正文

Building on SGLang's serving framework and AMD's MoRI communication library, we demonstrate that AMD achieves competitive — and at key operating points, superior — TCO compared to NVIDIA B200 running Dynamo + TRT-LLM. These results are validated by InferenceX, SemiAnalysis's open-source continuous benchmark platform that tests across hundreds of GPUs with a live dashboard. This post describes what we achieve, how we achieve it, and our plans for the road ahead. TL;DR At 129 tok/s/user interactiv

匹配字段正文