H200
PulseAugur coverage of H200 — every cluster mentioning H200 across labs, papers, and developer communities, ranked by signal.
- competes with Tiangai 300 90%
- instance of Blackwell 90%
- instance of graphics processing unit 90%
- used by NVIDIA H100 70%
- instance of NVIDIA H100 70%
- competes with Blackwell 70%
- used by GLM 5.3 70%
- competes with GLM 5.3 70%
- competes with RTX 5090 70%
- used by RTX 5090 70%
- acquired by Blackwell 70%
- used by Supermicro 70%
- 2026-08-19 product_launch First shipments of Nvidia's H200 AI chips have reached China, with ByteDance and Tencent as initial recipients. source
- 2026-05-14 product_launch US government approves sale of NVIDIA H200 AI chips to ten Chinese companies.
23 day(s) with sentiment data
-
Clover-35B-A3B diffusion model integrated with llama.cpp for local use
The Clover-35B-A3B-Preview model, a diffusion language model, has been integrated into llama.cpp, a popular tool for running models locally. Initial tests on an RTX 4090 showed a speed of 202-211 tokens per second, with…
-
Nvidia chips banned from China found in top Chinese AI models · 1 source tracked
A new research effort claims that China's most advanced AI models were trained using Nvidia chips that are banned from export to the country. The study suggests that these models, including those from DeepSeek and Moons…
-
Guide details cheap Qwen model fine-tuning with LoRA and QLoRA
This article provides a technical guide on efficiently fine-tuning Qwen models, particularly focusing on the use of LoRA and QLoRA techniques for cost reduction. It details how to select appropriate GPU hardware based o…
-
US lawmakers probe Trump admin's Nvidia chip sales to China
US lawmakers Richard Neal and Ron Wyden have launched an investigation into the Trump administration's handling of Nvidia's H200 chip sales to China. The inquiry focuses on a "scheme" involving tariffs and Commerce Depa…
-
Democrats probe Trump admin's Nvidia chip sales to China
Democrats have launched an investigation into the Trump administration's handling of Nvidia's chip sales to China, prompted by allegations of an illegal export tax scheme. The inquiry, led by Senators Ron Wyden and Repr…
-
Nvidia CEO Jensen Huang's alleged associate, Alice Huang, resurfaces in Beijing amid spy probe
Alice Huang, reportedly involved in an affair with Nvidia CEO Jensen Huang, has resurfaced in Beijing. Huang, suspected of being a Chinese spy, was previously detained in Singapore in 2025 as part of an investigation in…
-
Chinese university buys Nvidia H200 AI chips despite US ban · 2 sources tracked
A Chinese university with defense ties has reportedly purchased two Nvidia H200 AI chips, despite U.S. export restrictions. Northwestern Polytechnical University acquired the chips for approximately $123,000 USD through…
-
Chinese university buys Nvidia H200 AI chips, bypassing US ban
A Chinese university has reportedly purchased Nvidia H200 AI chips, circumventing US export restrictions. This acquisition is noted as a significant deal, potentially more valuable than previous arrangements involving O…
-
New LLM techniques: Byte-level conversion and efficient decoding strategies
Researchers have developed a method to convert a 27B parameter hybrid-attention LLM, Qwen3.8, into a pure byte-level model. This byte-level model, trained on self-generated data, shows competitive performance on several…
-
Trump: FBI investigating Nvidia CEO Jensen Huang's affair with suspected spy
President Donald Trump has stated that the FBI is investigating Nvidia CEO Jensen Huang's affair with a woman suspected of being a Chinese spy. This revelation comes amid separate allegations that Nvidia interfered with…
-
Mistral AI releases 'Le Chonk' 1T-parameter model for self-hosting
Mistral AI has released "Le Chonk," a 1 trillion parameter, multimodal model with a 1 million token context window, available for download. The model is offered under a modified MIT license that permits commercial use a…
-
Chinese Firms Approved for Nvidia AI Chips Linked to Military Suppliers, Investigation Finds
An investigation has uncovered that two Chinese companies, approved to purchase banned Nvidia H200 AI chips, appear to be shell corporations with empty offices and unknown addresses. These companies, New Matrix and Magi…
-
China blocks 60,000 Nvidia H200 AI chips, forcing domestic reliance
China's own regulators have reportedly blocked the import of approximately 60,000 Nvidia H200 AI accelerators, stranding them in Hong Kong. This move is seen as an effort by Beijing to bolster its domestic AI hardware i…
-
LLM context length limited by hardware memory, not model size
A recent analysis reveals that the ability of large language models to handle long contexts is primarily determined by the hardware's memory capacity rather than the model's architecture or quantization methods. The cri…
-
Self-hosting LLMs to cost $150K-$2M+ annually in 2026
Self-hosting open-source large language models in 2026 is projected to cost between $150,000 and over $2 million annually. The significant expenses extend beyond GPU hardware to include the cost of specialized engineeri…
-
LLM VRAM Needs: Llama, Qwen, and DeepSeek Self-Hosting Guide
Self-hosting large language models requires careful consideration of VRAM needs, which are influenced by model weights and KV cache size. Dense models like Llama scale predictably, with the 70B version demanding signifi…
-
Cost analysis reveals fine-tuning 70B models ranges from $1-2 per million tokens
Fine-tuning a 70 billion parameter model can cost between $1 and $2 per million training tokens, with a 20 million token LoRA run estimated at around $20. The cost is primarily determined by the number of tokens trained…
-
Chinese AI lab Moonshot seeks Nvidia GPUs in Europe to bypass US export bans
Chinese AI lab Moonshot, known for its Kimi model, is reportedly seeking to lease Nvidia GPUs in European data centers to circumvent U.S. export restrictions. The company is primarily interested in inferencing capacity …
-
HyperFlow H3 Flash model generates video and audio with fewer steps
Video Rebirth has released HyperFlow H3 Flash, a compressed version of MiniMax-H3 that generates synchronized video and stereo audio. This model, available in 14B parameter sizes, can produce 5.2 to 15-second videos at …
-
Nvidia struggles to meet China's $70B AI chip demand amid supply woes
Nvidia is reportedly unable to meet the massive demand for its RTX Pro 5500 AI chips in China, with sources indicating only about 1% of the estimated $70 billion order will be fulfilled in the coming months. Despite the…