NVIDIA has released TensorRT Model Connect, an open collection of reference implementations that enables developers to deploy open AI models from a Hugging Face model ID to native C++ inference in two…
NVIDIA introduced NVLink Fusion and NVHBM, a custom HBM base-die technology validated with leading memory vendors, to help hyperscalers and AI-native companies deploy custom XPUs into NVIDIA's AI infr…
NVIDIA published a tutorial on training a cross-embodiment robot navigation policy using an agent-driven workflow based on its COMPASS framework, which reuses the pretrained NVIDIA X-Mobility policy a…
Alibaba released the model weights for Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 176B total parameters, as a preview of the upcoming Qwen4 architecture. NVIDIA provides Day 0 func…
Alibaba released the model weights for Qwen3.8-Flash-Next, a multimodal mixture-of-experts model with 125B parameters plus 51B N-gram embeddings and 6B activated per token, as a preview of the Qwen4 a…
NVIDIA introduced shadow engine recovery, a preview feature in NVIDIA Dynamo, that restores LLM inference capacity in seconds by keeping a fully initialized shadow engine idle on the same GPUs, using …
NVIDIA released CUDA Python 1.0 with CUDA 13.3, making Python a supported, first-class way to use the CUDA platform, with components including cuda.core 1.0.0, cuda.compute 1.0.0, cuda.bindings 13.3.0…
NVIDIA's Spectrum-X Ethernet architecture, designed for giga-scale AI factories, maintains stable training step times of 668 ms under heavy multi-tenant congestion, while standard Ethernet slows from …
SemiAnalysis's AgentX benchmark shows NVIDIA's Vera Rubin NVL72 delivers up to 30x higher AI-factory throughput per megawatt than GB300 NVL72 for agentic-coding inference, while GB300 NVL72 extends it…
NVIDIA introduced Scale-In network infrastructure, the fifth pillar of its AI networking, powered by the NVIDIA BlueField-4 DPU and NVIDIA DOCA, to accelerate security, data movement, and operations f…
NVIDIA introduced the Vera CPU, designed to handle the unpredictable and highly variable workloads of agentic AI fleets, based on telemetry from 163,594 agentic sessions showing over 97% unique trajec…
NVIDIA announced that its Groq 3 LPX inference accelerator, paired with the NVIDIA Vera Rubin NVL72 platform, achieved a world-class interactivity of 3,431 output tokens per second on the Gemma 4 31B …
NVIDIA introduced DSX MaxLPS, a suite of chip, thermal, system, and software technologies designed to maximize AI factory throughput within a fixed power budget, addressing the challenge that only abo…
NVIDIA's AI safety and security teams outlined where security should live in an AI agent stack, emphasizing that infrastructure controls, not behavioral controls, are authoritative in determining what…
NVIDIA's Agentic Variation Operators (AVO) architecture achieved a 100% score on the ARC-AGI-3 benchmark, demonstrating a frontier-level general-purpose architecture for long-horizon autonomous agents…
NVIDIA's recsys-examples and nv-embedding-cache address the production challenges of generative recommenders (GRs), which reframe recommendation as sequence modeling to handle terabyte-scale user hist…
NVIDIA has demonstrated that its Holoscan platform can be used with AI coding agents to build real-time edge AI applications, using an endoscopic tool segmentation app as an example. The workflow, whi…
NVIDIA FLARE, an open-source federated learning framework, enables federated training of vision-language models (VLMs) across distributed sites by coordinating model updates through externalization, t…
NVIDIA released Cosmos 3 Edge, a 4B omni-model with a 2B Nemotron-based reasoner, designed for on-device robot control on Jetson Thor. Post-trained on the nvidia/Cosmos3-DROID dataset (76k trajectorie…
NVIDIA released SkillEvaluator, an open-source tool that measures how AI agent skills affect performance, and published benchmark results for more than 300 verified skills across over 30 NVIDIA produc…