Tag
据Axios报道，获NVIDIA领投20亿美元融资的美国AI初创公司Reflection AI即将发布一款开源模型，目标对标DeepSeek与Qwen，被视为美国在开源大模型领域对中国的一次重要回应。
NVIDIA introduces Sigma, the first large-scale (3B/8B) continuous diffusion language model, which denoises Gaussian-corrupted token embeddings over steerable low-dimensional ODE/SDE trajectories, achieving performance parity with discrete diffusion and autoregressive models on math and coding benchmarks while enabling embedding-space steering for quality-diversity trade-offs and efficient few-step distillation.
NVIDIA Labs released PixelUMM, an encoder-free unified multimodal model that operates directly in raw pixel space for both image and video understanding and generation, along with open-sourced code and models.
A tweet noting NVIDIA's stock hit a new all-time high and commenting on how AI-generated videos on Bilibili have become virtually indistinguishable from real content.
A community-curated list of budget and legacy GPUs (NVIDIA GTX/Maxwell/Pascal, AMD Radeon and Instinct MI-series) that can run LLM inference through the Vulkan backend on Linux, bypassing the latest CUDA/ROCm support requirements. The AMD Instinct MI50 16GB at around $150 is highlighted as the top value pick.
Jensen Huang shares his view that in the age of AI, subject choice matters less than storytelling, arts, and knowing what questions to ask, and stresses using AI to elevate learning and craft.
Strata 是一个免费开源的推理引擎，可在 RTX 4090 等消费级显卡（12GB+ 显存）本地运行 Qwen3.8-Flash-Next（125B）模型，写入速度约 100 tokens/s，且支持 Windows/Linux 和 NVIDIA/AMD 显卡。
NVIDIA开源了名为Lyra 2.0的模型，可将任意单张图片直接转换为可自由探索的交互式3D世界，甚至可用于向生成的场景中放入机器人进行仿真训练。模型已发布到Hugging Face，UI代码也已在GitHub开源。
'Big Short' investor Michael Burry is raising his timeline for an AI market crash, expanding short positions with put options on Nvidia, Palantir, Oracle, and the Nasdaq 100, arguing that heavy AI spending and circular dealmaking could cause the bubble to burst within a year.
A tweet highlights a free ~5% tokens-per-second performance boost on Blackwell GPUs by enabling CustomAllReduce on SM120 when using tensor parallelism 4, joking that closed labs hate this trick.
A post details the booming underground GPU repair industry in Shenzhen, where technicians service banned NVIDIA H100/A100 cards, alongside reports that China's H20 demand far outstrips NVIDIA's inventory amid ongoing US export restrictions.
NVIDIA CEO Jensen Huang shared a lighthearted personal anecdote while speaking at The Korea Society Annual Gala in New York on September 28.
OpenAI reportedly explored a $100M investment in Hugging Face before Nvidia's deal, with the platform potentially helping distribute OpenAI's own chips, raising questions about Hugging Face's neutrality under a chipmaker.
A tweet commenting on why NVIDIA DGX Spark prices are rising, linking to an external resource on the topic.
The US arrested Earthmade Computer CEO Greg Lui for allegedly smuggling over $300 million in export-controlled Nvidia A100 and H100 GPU servers into China through Malaysia and Singapore, part of a broader pattern of Nvidia-related chip-smuggling cases.
New teasers suggest NVIDIA's RTX Spark lineup of laptops and mini desktops will launch on October 7th, with rumored variants ranging from 24GB to 128GB of memory for local AI workloads.
A user shares spending $60,000 on 12 DGX Sparks to run near-frontier AI locally, reacting to news that NVIDIA has released a new 64GB DGX Spark at a higher price across all models.
Nvidia raised the price of the seven-year-old Shield TV Pro by $100 to $299.99, citing surging component and memory costs driven by the AI industry's demand for memory and storage.
NVIDIA is introducing a 64GB configuration of the DGX Spark personal AI supercomputer, sold through partners like Dell, HP, and ASUS, with clustering support via NVIDIA Sync to pool memory up to 128GB for running models up to 200B parameters locally.
OpenAI's GPT-6 Astra Ultrafast mode, running on NVIDIA Blackwell GPUs, is now available via the OpenAI API and to eligible ChatGPT Work/Codex users, delivering up to 8x faster token generation than Astra Standard mode. OpenAI reports using its own models to optimize inference kernels on NVIDIA hardware, improving latency, throughput, and cost for agentic coding workflows.