Inference

Introducing NSA: A Hardware-Aligned, Natively Trainable Sparse Attention Mechanism for Ultra-Fast Long-Context Training and Inference by DeepSeek AI

Introducing NSA: A Hardware-Aligned, Natively Trainable Sparse Attention Mechanism for Ultra-Fast Long-Context Training and Inference by DeepSeek AI

ByDeepMind April 30, 2025 4:18 pm

Researchers from Google DeepMind Unveil InfAlign: A Framework for Aligning Language Models with Inference Awareness

Researchers from Google DeepMind Unveil InfAlign: A Framework for Aligning Language Models with Inference Awareness

ByDeepMind April 24, 2025 7:40 am

AMD Foresees Mobile and Laptop Inference as the Future, Sees Potential to Compete with NVIDIA's AI Leadership

AMD Foresees Mobile and Laptop Inference as the Future, Sees Potential to Compete with NVIDIA’s AI Leadership

ByDeepMind April 21, 2025 1:10 amApril 21, 2025 1:10 am

AMD Explores New Possibilities as AI Workloads Shift to Inference

AMD Explores New Possibilities as AI Workloads Shift to Inference

ByDeepMind April 19, 2025 2:36 pmApril 19, 2025 2:36 pm

Implement DeepSeek-R1 Distilled Models on Amazon SageMaker with a Large Model Inference Container

Implement DeepSeek-R1 Distilled Models on Amazon SageMaker with a Large Model Inference Container

ByDeepMind April 17, 2025 12:42 amApril 17, 2025 12:43 am

Introducing DeepSeek V3-0324: Experience the World's Fastest Inference on SambaNova Cloud

Introducing DeepSeek V3-0324: Experience the World’s Fastest Inference on SambaNova Cloud

ByDeepMind April 13, 2025 6:42 amApril 13, 2025 6:43 am

The inaugural Google TPU for the era of inference

The inaugural Google TPU for the era of inference

ByDeepMind April 9, 2025 9:03 pmApril 9, 2025 9:03 pm

DeepMind Introduces Inference Time Scaling for Diffusion Models

DeepMind Introduces Inference Time Scaling for Diffusion Models

ByDeepMind April 8, 2025 1:31 pmApril 8, 2025 1:31 pm

DeepSeek Disrupts AI Field: The Next Breakthrough in AI May Rely on Enhanced Computational Power at Inference Instead of Increased Data

DeepSeek Disrupts AI Field: The Next Breakthrough in AI May Rely on Enhanced Computational Power at Inference Instead of Increased Data

ByDeepMind April 6, 2025 7:33 pmApril 6, 2025 7:33 pm

Supermicro Launches First NVIDIA HGX™ B200 Systems, Showcasing AI Performance Leadership in MLPerf® Inference v5.0 Results

Supermicro Launches First NVIDIA HGX™ B200 Systems, Showcasing AI Performance Leadership in MLPerf® Inference v5.0 Results

ByDeepMind April 3, 2025 6:59 pmApril 3, 2025 6:59 pm

NVIDIA Shifts Its Attention to Inference at GTC Following DeepSeek

NVIDIA Shifts Its Attention to Inference at GTC Following DeepSeek

ByDeepMind March 19, 2025 5:10 amMarch 19, 2025 5:10 am

Companies Invest Billions in Inference Rather Than Model Training

Companies Invest Billions in Inference Rather Than Model Training

ByDeepMind March 13, 2025 5:36 pmMarch 13, 2025 5:36 pm