Nvidia Blackwell B200
PulseAugur coverage of Nvidia Blackwell B200 — every cluster mentioning Nvidia Blackwell B200 across labs, papers, and developer communities, ranked by signal.
- instance of Blackwell 90%
- instance of GB200 90%
- instance of Nvidia Hopper 90%
- instance of Nvidia B200 90%
- competes with jalapeño 90%
- used by NVIDIA Dynamo 90%
- competes with AMD Radeon AI Pro R9700 80%
- competes with Intel Arc Pro B70 80%
- used by vLLM 70%
- used by Nvidia Hopper 70%
- used by NVFP4 70%
- instance of NVFP4 70%
7 day(s) with sentiment data
-
Fireworks AI launches new products to boost agent speed and cut costs
Fireworks AI has launched a suite of enterprise products aimed at improving AI agent performance and cost-efficiency. The company demonstrated speeds of up to 1,500 tokens per second on its DeepSeek V4.1 Flash model, hi…
-
FLAMIA kernel accelerates Transformer training on NVIDIA Blackwell GPUs
Researchers have developed FLAMIA, a new attention kernel designed to accelerate Transformer training on NVIDIA Blackwell GPUs. FLAMIA utilizes mixed-precision kernels, specifically MXFP8 and NVFP4, to reduce quantizati…
-
Nvidia releases quantized DeepSeek-V4.1-Flash-1017; uncensored version also emerges
Nvidia has released a quantized version of DeepSeek AI's DeepSeek-V4.1-Flash-1017 model, optimized for commercial applications and featuring a multimodal Mixture-of-Experts architecture with Compressed Sparse Attention …
-
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…
-
MiniMax M3.1 Flash cuts AI agent costs by 70% with 428B MoE model
MiniMax has released its M3.1 Flash model, a 428-billion-parameter Mixture-of-Experts (MoE) architecture designed to significantly reduce costs for agentic AI applications. The model achieves this by activating only a f…
-
MiniMax M3.1 Flash open weights released with vLLM day-zero hardware support
MiniMax has released the open weights for its M3.1 Flash model, a multimodal mixture-of-experts architecture with 428 billion total parameters and a 1 million token context window. The model utilizes MiniMax Sparse Atte…
-
Modal and Sakana AI deploy trillion-parameter Japanese LLM with custom quantization
Modal has partnered with Sakana AI to deploy Sakana Namazu, a trillion-parameter Mixture-of-Experts (MoE) model optimized for Japanese business contexts, web search, and code execution. The collaboration focused on over…
-
DeepSeek open-sources DeepInfer for Nvidia and Huawei GPUs
DeepSeek has released DeepInfer, an open-source inference framework aimed at optimizing AI model performance on Nvidia Blackwell and Huawei GPUs. This move is positioned as a strategic effort to counter closed-source AI…
-
New Token-Energy Law quantifies LLM decoding energy use
Researchers have introduced the Token-Energy Law, a model that quantizes the energy consumption of large language models (LLMs) during decoding. This law relates energy per token to "active bytes" (logical byte demand) …
-
Frore Systems unveils diamond-infused cooling for Nvidia Rubin AI accelerators
Frore Systems has developed a new liquid-cooling solution called LiquidJet Diamond, designed for high-power AI accelerators like Nvidia's Rubin and AMD's Instinct MI400/MI500 series. This bespoke coldplate technology in…
-
MiniMax AI open-sources faster VC Attention for Nvidia Blackwell B200
MiniMax AI has open-sourced a variant of VC Attention optimized for Nvidia Blackwell B200 GPUs. This new attention mechanism reportedly achieves approximately double the speed of BF16 FlashAttention on the B200, specifi…
-
MiniMax AI open-sources faster Nvidia Blackwell attention variant
MiniMax AI has open-sourced a variant of VC Attention optimized for Nvidia Blackwell B200 hardware. This new implementation reportedly achieves approximately double the speed of BF16 FlashAttention on the B200, offering…
-
Prime Intellect launches Prime Inference serving platform for open models
Prime Intellect has launched Prime Inference, a new serving platform designed for frontier open-source AI models. The platform offers both serverless endpoints for variable demand and reserved capacity for sustained wor…
-
New polynomial approximations boost LLM training speed on NVIDIA Blackwell GPUs
Researchers have developed new polynomial approximations for transcendental functions used in large language models (LLMs) to improve computational efficiency. These approximations, tested on NVIDIA Blackwell GPUs, show…
-
New SPARC method enhances sequence model efficiency and training speed
Researchers have developed a new method called Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence (SPARC) to improve sequence models. SPARC uses minimal input-dependent signals to manage memor…
-
New SPARC method enhances sequence models with efficient adaptive spectral memory
Researchers have developed a new method called Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence (SPARC) to improve sequence models. SPARC uses minimal input-dependent signals to manage memor…
-
AI Hardware Race Intensifies: Nvidia Faces Challenges from AMD, Intel, and Cloud Giants
The AI hardware landscape is intensely competitive, with Nvidia currently leading the pack. However, rivals like AMD and Intel are making significant strides, and major cloud providers such as Microsoft, Google, and Ama…
-
UBTECH opens massive humanoid factory, aiming for 10,000 units annually
UBTECH Robotics has opened a new factory in China capable of producing 10,000 humanoid robots annually, with a new unit rolling off the assembly line every 10 minutes. This significant scaling effort places UBTECH along…
-
NVIDIA boosts AI factory efficiency with new platforms and partnerships
NVIDIA announced advancements in its AI infrastructure platform at the AI Infra Summit, focusing on energy efficiency and token optimization. Key developments include collaborations with Amazon's Annapurna Labs on high-…
-
NVIDIA powers AI robotics and robotaxis with advanced physical AI platforms
NVIDIA is collaborating with Skild AI to advance robotics by enabling robots to learn new tasks from single video demonstrations, a technique known as in-context learning. This approach, powered by NVIDIA's AI infrastru…