NVIDIA GH200
PulseAugur coverage of NVIDIA GH200 — every cluster mentioning NVIDIA GH200 across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
Researchers hack AI hardware to make chatbots give dangerous advice
Researchers at Eclypsium have demonstrated a method to rewrite AI hardware weights in memory, specifically targeting an Nvidia GH200 server. This exploit, performed in under a minute and bypassing standard security chec…
-
FlashSO(2) framework accelerates materials modeling with low-precision emulation
Researchers have developed FlashSO(2), a novel framework designed to accelerate the training and inference of SO(2)-equivariant graph neural networks used in materials modeling. This framework utilizes low-precision emu…
-
Developer shares lessons learned building APEx 2-hybrid LLM
A developer shared lessons learned from building the APEx 2-hybrid model, a quantitative protein-protein interaction assay for antibody discovery. The primary bottleneck was GPU availability, leading to a reduced traini…
-
New AI model LAVOIR learns to ask the right questions
Researchers have developed LAVOIR, a novel decision encoder designed to improve how AI systems ask clarifying questions. Unlike previous models that guess when information is missing, LAVOIR identifies and prioritizes t…
-
AI model BEAST achieves high-fidelity atmospheric forecasting with novel 4D-parallelism
Researchers have developed BEAST, a Bayesian Swin Transformer model for atmospheric forecasting that can quantify uncertainty. This model utilizes a novel 4D-parallelization scheme to efficiently train a 2.4-billion-par…
-
Qwen3.8-27B open-source model rivals Claude Opus on benchmarks
Qwen has released its Qwen3.8-27B model, an open-source model that rivals the performance of larger, proprietary models like Claude Opus 4.6 Max on various benchmarks, particularly in coding and agent tasks. This 27-bil…
-
DSv4-Flash optimization boosts LLM inference speed on NVIDIA GH200
A new optimization technique called DSv4-Flash has been developed to significantly speed up large language model inference on NVIDIA GH200 hardware. This optimization, when implemented with vLLM and SGLang, can achieve …
-
GRACE system accelerates real-time ad retrieval with generative recommenders
A new research paper introduces GRACE, a system designed to accelerate generative recommenders for real-time ad retrieval. GRACE addresses challenges in eligibility and compute by implementing Generative Target Matching…
-
TESSERA v2 study reveals optimal scaling for Earth-observation models
Researchers have conducted a large-scale study on scaling pixel-wise Earth-observation foundation models, involving 395 training runs on 1,024 NVIDIA GH200 superchips. The study found that pretraining loss is a poor pre…
-
GLM-5.2 model speed boosted over 20x via custom hacks
A Reddit user detailed a method for significantly accelerating the GLM-5.2 large language model on a specialized GH200 system. By combining components from different repositories and patching the vLLM inference engine, …
-
DiffusionGemma 26B on GH200 shows extreme speed, 32K context handling
A technical deep-dive reveals that the DiffusionGemma 26B model, when run on NVIDIA's GH200 Grace Hopper platform with vLLM optimization, achieves exceptional performance. The setup demonstrated a generation throughput …