H100s
PulseAugur coverage of H100s — every cluster mentioning H100s across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
Gemma LLM praised for local deployment economics and developer experience
The author highlights the practical economic advantages of using smaller, locally deployable Large Language Models (LLMs) like Google's Gemma, particularly for developers in regions like Africa. The piece emphasizes tha…
-
OpenAI unveils custom chip, DeepSeek boosts LLM speed, local models degrade
OpenAI has developed a custom inference chip codenamed Jalapeño, in collaboration with Broadcom, designed specifically for efficient LLM operation. This move aims to reduce reliance on NVIDIA and potentially lower API c…
-
Cursor AI launches custom model trained on 1M H100s, introduces Origin platform
The creators of the Cursor code editor have announced their own AI model, trained on one million H100 GPUs. They have also launched a new platform called Origin, designed to simultaneously manage thousands of AI agents.…
-
Groq's custom LPU chip offers 10x memory bandwidth for faster LLM inference
Groq has developed a novel Language Processing Unit (LPU) that significantly outperforms traditional GPUs for large language model (LLM) inference. Unlike GPUs, which were designed for graphics and repurposed for AI tra…
-
Ohio State researchers open-source QUEST-35B Deep Research agent
Researchers from Ohio State University have developed and open-sourced QUEST-35B, a Deep Research agent. This agent was trained using approximately 32 H100 GPUs and a dataset of around 8,000 synthetic samples. Benchmark…
-
AI GPU shortage questioned amid findings of significant underutilization
A recent analysis of GPU utilization in AI workloads suggests that the perceived shortage of high-end GPUs like NVIDIA's H100s and Blackwell B200s may be exacerbated by underutilization. The GPU in question spent a sign…
-
Together Compute Expands GPU Offerings with H100, H200, and B200
Together, an inference and open-source AI company, has significantly expanded its on-demand compute platform. The company announced the addition of a substantial number of high-end GPUs, including H100s, H200s, and the …
-
Prefix Caching Slashes LLM Prefill Costs by 80%
A new technical article explores prefix caching as a method to significantly reduce the computational cost of processing long prompts in large language models. This technique is particularly effective for workloads like…
-
New tool DODOCO reveals flaws in MoE model dispatch benchmarks
A new research paper introduces DODOCO, a tool designed to diagnose overhead in dispatch operations for Mixture-of-Experts (MoE) models. The study found that common assumptions about workload representation in benchmark…