Nvidia H800
PulseAugur coverage of Nvidia H800 — every cluster mentioning Nvidia H800 across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
Moore Threads' 2,112-GPU Wuxi Cluster Hits 92.9% Commercial Utilization
Moore Threads has achieved high commercial utilization for its 2,112-GPU cluster located in Wuxi. The cluster is currently serving 55 customers and has reached a 92.9% utilization rate. Initial tests using DeepSeek V4 F…
-
DuplexTune framework offers memory-efficient adaptation for MoE language models
Researchers have developed DuplexTune, a new framework designed to make adapting Mixture-of-Experts (MoE) language models more memory-efficient. This method involves replaying coupled hidden states through pretrained mo…
-
Self-hosting LLMs costs more than APIs for bills under $10.5K/month
A recent analysis suggests that for API bills under approximately $10,500 per month, using cloud-based LLM APIs is more cost-effective than self-hosting models on dedicated hardware. The calculation considers the monthl…
-
New method compiles LLM control for faster, more accurate Hierarchical RAG
Researchers have developed a new method for improving the efficiency of large language models (LLMs) when used as controllers in structured decision pipelines, specifically within Hierarchical RAG systems. This techniqu…
-
AI firms explore personalized models for long-term user intelligence
Tongyan Lingzhi, a team incubated by the Beijing Academy of Artificial Intelligence (BIGAI), is developing a Recursive Self-Improving Personal AI (RSI) system. This system aims to allow individuals and organizations to …
-
AUGUstus VLA model unifies action generation and understanding for autonomous driving
Researchers have developed AUGUstus, a novel Vision-Language-Action (VLA) model designed for autonomous driving that unifies action generation and understanding to overcome limitations of existing VLA systems. Unlike pr…
-
AI agents evolve skills and policies across multiple research papers · 10 sources tracked
Recent research explores methods for improving the self-evolution capabilities of AI agents, particularly in complex environments. Papers introduce techniques like CO-evolution Drift Audit (CODA) to evaluate agent robus…
-
New scaling laws unlock hyperparameter tuning for ultra-sparse MoE models
Researchers have developed new hyperparameter scaling laws specifically for ultra-sparse Mixture-of-Experts (MoE) models, addressing challenges in transferring optimal learning rates and batch sizes across varying spars…
-
New Rollplex runtime boosts VLM post-training efficiency
Researchers have developed Rollplex, a new runtime system designed to improve the efficiency of post-training vision-language models (VLMs) using reinforcement learning. Current methods process rollout, reference scorin…
-
New methods emerge for efficient visual token pruning in AI models · 6 sources tracked
Researchers are developing new methods to optimize Vision Transformers (ViTs) and Multimodal Large Language Models (MLLMs) by pruning visual tokens, which are computationally expensive. Several papers propose novel tech…
-
New KV Cache Compression Techniques Boost LLM Inference Performance · 9 sources tracked
Multiple research papers explore novel techniques for optimizing the Key-Value (KV) cache in large language model (LLM) serving to address memory and performance bottlenecks. These methods, including quantization, pruni…
-
New QK-Normed MLA method stabilizes LLM attention without full key caching
Researchers have developed QK-Normed MLA, a method to stabilize attention mechanisms in large language models without requiring full key caching. This technique integrates QK normalization into Multi-head Latent Attenti…
-
China's Military Acquires Nvidia AI Chips Despite US Export Controls
Research indicates that China's military has continued to acquire advanced Nvidia AI chips, even after U.S. export controls were implemented. Publicly available documents reveal numerous procurement requests from variou…
-
ChunkFT framework slashes fine-tuning memory needs for Llama 3
Researchers have developed ChunkFT, a new framework designed to make full-parameter fine-tuning of large language models more memory-efficient. This method allows for gradient computation on dynamic subsets of model par…