Sakana AI has released Fugu Max and Fugu Ultra v2, 2 models built on the same learned orchestration architecture. Fugu Max routes tasks to lean open and specialized models, including NVIDIA Nemotron, at $2/$6 per 1M tokens. Fugu Ultra v2 targets peak capability, scoring 48.3 on Chartography and 74.3 on DeepSWE. The post Sakana AI Launches Fugu Max and Fugu Ultra v2 for Cheaper, Stronger Multi-Agent Orchestration appeared first on MarkTechPost.
arXiv:2609.10951v1 Announce Type: new Abstract: AI-based automated vehicle testing is challenging because a model that passes every test condition can still fail in the real world. Formal verification offers a way to directly address this gap. On a simulated highway and an arterial road we trained two small end-to-end steering networks each in CARLA, one on clear conditions alone and one on clear, fog, night and low sun. All four models were driven against a 2.19 ft lane-departure budget. Without driving again, we used bound propagation, a formal method that reads the trained weights, to compute how far steering can drift at every disturbance strength between two captured images. One calculation covers more than a campaign could drive: on the arterial it spans 133 poses, where ten intensi…
NVIDIA has detailed BioNeMo Inference Runtime (BioIR), a Python library that accelerates biomolecular structure-prediction models on NVIDIA GPUs while staying in plain PyTorch. In a matched benchmark on 1,000 human dimer targets across 8xH100 GPUs, BioIR-accelerated Boltz-2 delivered 58.5K successfully folded residues per GPU-hour versus 20.2K for a torch-compiled open-source implementation, a 2.90x gain. The runtime optimizes at 3 layers: custom kernel selection, CUDA Graph capture, and Ray-based replica scaling that places 1 full model copy per GPU. BioIR already powered the AlphaFold Database expansion, generating about 31 million candidate protein complexes across 4,777 proteomes. The post NVIDIA Details BioNeMo Inference Runtime (BioIR): 2.90x Higher Boltz-2 Folding Throughput and 58…
Amazon SageMaker HyperPod now supports model caching for inference, which pre-loads model weights and container images onto cluster nodes so pods read from local NVMe storage instead of downloading over the network. Learn how model caching cuts cold starts from tens of minutes to seconds, how it works, and how to enable it.
Red Hat released Red Hat AI 3.5 this week, a move designed to let software engineering teams run AI with The post Red Hat AI 3.5 tackles the GPU queue that can stall AI pilots appeared first on The New Stack.
Manufacturing floors, warehouses and production lines rarely stay fixed — tasks change, layouts shift and new products arrive, and most robots can’t keep up without significant reprogramming. Skild AI’s new S1 robot foundation model helps address this, designed to learn previously unseen, long-horizon tasks from a single video demonstration. The model, launched last week, uses […]
The global robotaxi market — physical AI’s first commercial breakthrough — is projected to reach $400 billion by 2035, with over 6 million commercial vehicles in operation as driverless fleets are already moving people through some of the world’s busiest and most complex streets. Deploying a driverless vehicle is one challenge. Scaling a fleet is […]
Universal Music Group is launching a new AI-powered platform that will allow users to draw from its catalog of licensed music to create song remixes, mashups, and new takes on tracks, according to an announcement on Thursday. The record label is developing the platform through a multi-year licensing agreement with ElevenLabs, a company that specializes in AI voice and music generation. Artists can choose whether to participate in UMG and ElevenLabs' upcoming platform, which marks yet another AI deal for the record label. UMG is currently developing an AI music platform with Udio and has struck AI licensing deals with Spotify, Nvidia, and Kl … Read the full story at The Verge.
AI inference chipmaker d-Matrix today announced it will use NVIDIA NVLink Fusion to connect its next-generation Raptor XPUs to NVIDIA’s AI infrastructure platform — joining a growing roster of ecosystem partners. By connecting Raptor to NVIDIA NVLink scale-up and Spectrum-X scale-out networking, the NVIDIA MGX rack architecture and the broader NVIDIA AI platform, NVLink Fusion […]
Gear up: The latest PC games and major updates are ready to play on GeForce NOW this week. WARDOGS drops onto the cloud at early-access launch, alongside the Valheim 1.0 Deep North update and Bus Simulator 27 — part of nine new titles joining the cloud. The newest PC releases can demand serious hardware, storage […]
Nvidia and Palantir announced on Thursday that they’re working together to bring “sovereign AI to critical supply chains,” kicking off The post “AI factories are among the most complex systems ever built”: Nvidia and Palantir turn Nvidia’s supply chain into a proving ground for sovereign AI appeared first on The New Stack.
We ported ThunderKittens to NVIDIA's Vera Rubin NVL72 and rebuilt our NVFP4 GEMM around the new hardware, taking it from 42% of roofline to over 22 PFLOPS — competitive with cuBLAS and CuTe DSL. Here is what changed in the ISA and how we used it.
Learn how to deploy Qwen3.8-2.4T-A95B, a 2.4-trillion-parameter open-weight model, on Amazon SageMaker HyperPod with vLLM. This walkthrough covers cluster provisioning, NVFP4 quantization, and an OpenAI-compatible endpoint with built-in reasoning, tool calling, and native MTP speculative decoding.
At the IBC conference, running Sept. 11-14 in Amsterdam, the creative, technology and business communities are coming together to turn ideas into action and discuss innovations across the media and entertainment industries. More than 44,000 attendees from 170+ countries are gathering to explore 1,300+ exhibitions in 14+ halls and outdoor spaces, with over 600 speakers […]
TorchServe is no longer maintained, leaving teams to own the entire GPU inference stack. The AWS Ray Serve Deep Learning Container is a supported, pre-tested container with the framework, GPU drivers, and serving layer already assembled. This post walks through deploying a vision-language model on Amazon EKS using the Ray Serve DLC on a single GPU node.
Perplexity's engineering team published an in-depth look at the GPU serving infrastructure behind pplx-embed. The system reuses LLM prefill/decode kernels, separates responsibilities into a Rust gateway (Ivy), a gRPC server (Tulip), and a Python inference engine (ROSE), and uses CUDA graphs, lazy capture, and LazyTensor to optimize throughput and latency. Benchmarks against vLLM are also covered.
Nous Research has collapsed local model setup into a single click in Hermes Desktop. The app reads your hardware, fit-checks the catalog against your GPU, picks the highest-quality build that fits, downloads it, and configures llama.cpp — with a hard 4-bit floor and a 64K minimum context window.
Nvidia’s $12.9 billion Hugging Face acquisition extends the AI chip giant’s reach beyond compute, but preserving the platform’s openness will be key to its value.
Microsoft has officially named its developer-optimized Windows experience Project Zenith. The initiative targets new developer devices with 64GB or more of unified memory, allowing local and unmetered execution of 30B+ parameter models. AMD unveiled the first Project Zenith mini PC at IFA, and Microsoft confirms more devices with different silicon are coming. Preinstalled tools and default Windows tweaks aim to reduce interruptions.
Ugreen’s new HomeAgent platform combines local security-camera storage, on-device AI, and smart-home control in one hub, managed by the Uliya voice assistant. Designed to keep data in your home, it promises no subscription fees, with models ranging from a basic hub to an Nvidia-powered MasterAgent.
While most attention has been on model makers such as Anthropic, OpenAI and Google, Nvidia is emerging as the star of the AI race through its acquisition of Hugging Face.
Nvidia has agreed to acquire Hugging Face, the open platform often called the “GitHub of AI”, for roughly $12.9 billion. Nvidia pledged that Hugging Face will remain open, multi-cloud, and multi-accelerator, and that Nvidia compute will not be required to use it. The deal is expected to close in the first half of 2027, subject to regulatory approvals.
Nvidia's new open-source PAIR router routes local AI model requests to idle Macs and PCs in your home network, accelerating agentic workflows with subagents while working alongside Ollama or LM Studio.
Nvidia has introduced PAIR (Personal AI Router), free open-source software that discovers idle, compatible PCs on a home network and pools their power for local AI inference and agentic workloads. It supports Nvidia GeForce RTX 20-series and newer, RTX Pro GPUs, DGX Spark systems, and Apple M4 or newer chips. The beta is available today for Windows, Linux, and macOS.
GeForce NOW is adding 26 games in September, led by NBA 2K27 with NVIDIA DLSS 5 3D-Guided Neural Rendering. The Blood of Dawnwalker and Onimusha: Way of the Sword also arrive at launch, with Ultimate members able to stream on RTX 5080-class cloud rigs.
Nvidia has agreed to acquire open-source AI platform Hugging Face for $12.93 billion. The platform, often called the 'GitHub of AI,' hosts a vast library of open-source models, datasets, and tools. The deal would give Nvidia a strategic position in the open-source AI ecosystem as it looks to defend its dominance in AI hardware, while open-source developers race to catch up with closed AI systems.
NVIDIA has agreed to acquire Hugging Face for $12,930,300,000, planning to scale the platform while keeping it open and multi-cloud. The deal underscores NVIDIA's support for open-weight models and expands its role in the AI ecosystem.
New York City Mayor Zohran Mamdani announced a one-year moratorium starting in the 2026-2027 school year that bars about 600,000 public school students in 2-K through eighth grade from using AI in classrooms. Teachers can no longer use AI to grade assignments, companion and mental-health chatbots are banned in all grades, and high schoolers face limits plus AI literacy classes. The policy also restricts personal screens and launches a small pilot with vetted AI tools in high schools.
This paper introduces ZimaBlue, a scalable framework for learning generalizable World Action Models (WAMs) from large-scale video. It employs a three-stage training curriculum: causal embodied video pre-training on large-scale human and robot egocentric videos, video-action mid-training with unified action representation, and robot-specific specialization. Its asynchronous Slow-Fast dual-system architecture enables 30 Hz real-time action prediction on an NVIDIA RTX 4090. In real-robot zero-shot evaluations, scaling from target-robot data alone to over 120,000 hours of embodied video improves success from 36.1% to 77.8%, with particularly strong gains on unseen tasks.
CUDA-Harness is a new framework for generating and optimizing high-performance CUDA kernels from natural language. It introduces Intermediate-Structured Generation to connect high-level semantics with low-level kernel generation, uses Synthesis-Based Verification to mitigate reward hacking, and proposes Feedback-Adaptive Evolution to prioritize correctness while optimizing performance. Experiments show its effectiveness and generalization across LLMs, hardware, and C-to-CUDA transpilation.
At CrowdStrike's Fal.Con 2026, NVIDIA CEO Jensen Huang and CrowdStrike CEO George Kurtz unveiled SafeMind, an agentic cybersecurity system that uses NVIDIA Nemotron models and CrowdStrike's threat data to enable a continuous coevolution of offensive and defensive AI, aiming to defend against increasingly automated attacks.
With AI inference driving demand for compute, several companies are now tapping into idle computing power in homes and small businesses, paying device owners to run AI models. This distributed approach could be cheaper, more reliable, and avoid the downsides of massive data centers. Features Far Labs and Evolving Edge, among others, and discusses security measures, challenges, and potential new use cases.
shaide is a self-hosted AI platform that runs on your own Kubernetes clusters, designed for distributed multi-model inference at scale. It installs with a single command, stays entirely within your network perimeter, and even supports fully air-gapped environments. The platform manages its infrastructure as code with Pulumi, offers an OpenAI-compatible API, and is optimized for agent fleets.
Nvidia officially launches DLSS 5 this week, but only NBA 2K27 supports it, and it demands top-tier performance, requiring even a mid-range RTX 5060 to use 6x frame generation for 1080p gaming.
This article details NVIDIA's Nemotron 3 Ultra, a 550B-parameter hybrid Mamba-attention MoE model released on June 4, 2026. It activates only about 55B parameters per token (≈10% sparsity) and combines Mamba-2 state-space layers with transformer attention layers for long agentic runs. The post covers the Nemotron 3 family (Nano, Super, Ultra), the motivation for the hybrid architecture, training details, and how to call the model via OpenRouter, NVIDIA NIM, or self-hosted vLLM.
Private AI search across your work. Ask across Google Drive, Dropbox, and Slack. Get source-backed answers without a permanent index. Choose German hosting or your own infrastructure. Create free account Create your acc…
The Hidden Bottleneck: Why Advanced Packaging is the Next AI Moat Executive Summary The market's obsession with GPU design is noise. The true bottleneck—and strategic moat—in the AI hardware race is not the silicon itse…
We're excited to share that AWS has been recognized as a Leader in The Forrester Wave: AI Infrastructure Solutions, Q4 2025. In this evaluation of 13 providers, AWS received the highest score in the Strategy category.
Three-quarters of Americans said in a recent poll that they oppose datacenters being built next to their homes Donald Trump has criticized communities pushing back against datacenter projects across the US amid a growing backlash, warning those that reject them risk becoming “backwards and poor”. As controversy surrounding local datacenter plans continues to swirl around election campaigns nationwide ahead of November’s midterm elections, the US president declared Americans “will only have yourselves to blame” if they are canceled. Continue reading...
Investing Apple Is Suddenly an AI Infrastructure Stock as OpenAI Buys Macs by the Tens of Thousands OpenAI has been quietly buying Apple hardware by the tens of thousands, and it has nothing to do with iPhones or consum…
Today, I’m talking with New York Governor Kathy Hochul, and I’ll just warn you — this episode moves really fast. It’s an election year, after all, with a shocking amount of tech policy at stake, and Governor Hochul has taken strong positions on almost every major tech issue there is. For example, Meta just reached a settlement with dozens of states, including New York, which will restrict how teens use platforms like Instagram in very specific ways. Governor Hochul is a strong supporter of those restrictions and more, as you’ll hear. But those come with a cost — widespread age verification means adults will also have to show ID to use the internet, which will essentially make it impossible to be anonymous online. Verge subscribers, don’t forget you get exclusive access to ad-free Decoder…
arXiv:2608.27550v1 Announce Type: new Abstract: Scaling robot data is crucial for building generalist Vision-Language-Action (VLA) models, yet robot trajectories are harder to scale than web-scale image-text data because embodied collection is costly and sparsely covers the physical world. This makes representation quality a central bottleneck: under a fixed robot-data budget, continued pre-training must turn limited trajectories into transferable visual-action knowledge rather than merely fit actions. We propose VLAct, a VLA-oriented VLM backbone trained on broad, heterogeneous, multi-embodiment robot data before task-specific fine-tuning. VLAct preserves the broad VLM prior and encourages shared action semantics across embodiments through VLM-prior preservation, multi-head continuous ac…
arXiv:2608.27735v1 Announce Type: new Abstract: We present ABCD (Alpha-Composited Block Coordinate Descent), an out-of-core training framework for alpha-composited radiance fields, instantiated here for 3D Gaussian Splatting. Our method reformulates training as block coordinate descent over spatial partitions: only one block of parameters is active at a time, while all others are frozen. By exploiting the associativity of alpha blending, these inactive regions can be pre-rendered and collapsed into foreground and background RGBA images. As a result, for fixed partition size and image resolution, peak VRAM becomes O(1) with respect to total scene extent, rather than growing with full scene size. This enables GPUs with limited memory to train scenes that would otherwise not fit in core. In…
arXiv:2608.27584v1 Announce Type: new Abstract: Autonomous systems rely on extracting information from light, yet remain brittle in extreme environments, from nighttime navigation to high-speed robotics. Conventional sensors aggregate photons over fixed exposures, imposing trade-offs between sensitivity, dynamic range, and temporal resolution that degrade perception when photons are scarce or dynamics are rapid. Quanta sensors detect individual photons, but their streams exceed real-time compute and latency budgets by orders of magnitude. Here we introduce $\textit{probabilistic events}$, a computational primitive for real-time quanta perception from individual photon detections. By computing the posterior over the time since the last intensity change, we represent photon streams as recur…