NVIDIA H200
PulseAugur coverage of NVIDIA H200 — every cluster mentioning NVIDIA H200 across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
LLM power capping ineffective during decode, new research finds
A new research paper challenges the effectiveness of power capping on GPUs for Large Language Model (LLM) serving, particularly during the autoregressive decode phase. The study found that power caps have minimal impact…
-
MatAnyone-E achieves 7.4x speedup in high-resolution video matting
Researchers have developed MatAnyone-E, an efficient extension of MatAnyone 2 designed for high-resolution video matting. This new method focuses computational resources on uncertain regions, reducing the need for exten…
-
New leaderboard aims to speed up relational foundation model pretraining
Researchers have introduced modded-nanoRT, a new leaderboard designed to accelerate the pretraining of Relational Transformers. This system aims to remove barriers to foundation model pretraining by focusing on speed ra…
-
AI agents use hardware attestations for trustless settlement
A new settlement mechanism called Atomic Delegation aims to resolve payment issues in agent-based economies where tasks are subcontracted. Current methods involve intermediaries, leading to costs, risks, or delays. This…
-
Kolibri-1 (78B MoE) benchmarked for NIS-2 compliance on NVIDIA H200
A new benchmark test evaluated the Kolibri-1 (78B MoE) model on NVIDIA H200 hardware using vLLM, focusing on NIS-2 compliance, incident reporting, and jailbreak resilience. The tests revealed that Kolibri-1 successfully…
-
Fish Audio releases Drama 3 text-to-speech model with advanced direction features · 4 sources tracked
Fish Audio has released Drama 3, an open-source text-to-speech model with 35 billion parameters that activates 3 billion per token. This model offers advanced features such as multi-character scene generation, voice clo…
-
Physical side channels can identify AI workloads on NVIDIA H200 GPUs
Researchers have developed a method to identify AI workloads running on NVIDIA H200 GPUs by analyzing their power consumption, a technique that could be crucial for AI governance and policy enforcement. This approach us…
-
Research details serving challenges for faster diffusion language models
A new research paper on arXiv explores the challenges of serving masked diffusion language models (dLLMs), which can generate text faster than traditional autoregressive models by denoising multiple tokens simultaneousl…
-
Unitree Robotics stock surges 629% in Shanghai debut amid AI fervor · 4 sources tracked
Chinese robotics company Unitree Robotics experienced a remarkable stock market debut, with shares surging by up to 629% on their first day of trading on the Shanghai Stock Exchange. This significant jump highlights the…
-
Kog's new engine boosts AI inference speed on existing GPUs · 4 sources tracked
French startup Kog has developed a new inference engine designed to significantly accelerate AI model performance on existing datacenter GPUs, such as the AMD MI300X and NVIDIA H200. The company's software-driven approa…
-
AI compute startup B3IQ offers rent-to-own GPUs to university researchers
Startup B3IQ is addressing a unique market need by offering a rent-to-own model for high-performance AI computing hardware, specifically targeting university researchers. This approach provides academics with predictabl…
-
AI systems optimize GPU kernel performance for scientific computing
Researchers have developed two novel systems, SparseDitto and KernelBrain, aimed at optimizing GPU kernel performance for various computational tasks. SparseDitto utilizes an LLM-based agent to generate custom GPU kerne…
-
New method enhances manga image editing without retraining
Researchers have developed a new method for editing manga images that adapts existing image editing models without requiring retraining. This training-free approach modifies the editing trajectory to preserve the global…
-
Thinking Machines releases Inkling-Small, outperforming larger predecessor
Thinking Machines Lab has launched Inkling-Small, a new open-weights multimodal model that prioritizes efficiency over sheer size. Despite being significantly smaller than its predecessor, Inkling, Inkling-Small demonst…
-
Kimi K3 AI model optimizes GPUs, designs chips, and accelerates research
Kimi K3, an advanced AI model from Open Frontier Intelligence, has demonstrated remarkable capabilities beyond typical language tasks. The model has shown proficiency in optimizing GPU kernels, improving learning speeds…
-
GMO CEO apologizes for remote work ban, cites AI and office value
GMO Internet Group CEO Masatoshi Kumagai has apologized for his recent announcement to completely abolish remote work, clarifying that the company is not negating remote work itself but rather its policy of recommending…
-
Moonshot AI releases Kimi K3, challenging frontier AI models with open weights
Moonshot AI has announced Kimi K3, a new 2.8 trillion parameter open-weight model with a 1 million token context window, positioning it as a strong contender in the frontier AI space. While benchmarks suggest Kimi K3 pe…
-
Thinking Machines releases Inkling, an efficient multimodal MoE model
Thinking Machines has released Inkling, a new open-weight, multimodal Mixture-of-Experts model with 975 billion total parameters and 41 billion active parameters. The model supports a 1 million token context window and …
-
China eases Nvidia H200 chip ban for AI firms amid global race
China's government is reportedly easing restrictions on the purchase of NVIDIA H200 chips for select domestic AI companies, including Alibaba Group, ByteDance, and DeepSeek. This move signals a pragmatic approach to add…
-
NVIDIA releases Nemotron VoiceChat and Parse 2.0 models
NVIDIA has released two new models on Hugging Face: NVIDIA NemotronLabs VoiceChat 11B, an end-to-end, real-time speech model for conversational AI that supports full-duplex interaction and tool calling, and NVIDIA Nemot…