Records found: 4
Cerebras released MiniMax-M2-REAP-162B-A10B, a REAP-pruned SMoE checkpoint that trims experts to reduce memory while keeping near-original performance on code, reasoning and agentic tool calls.
Explore how optimizing AI inference can enhance performance, lower costs, boost privacy, and improve customer experience in real-time applications.
Researchers introduce Effective State-Size (ESS), a new metric that measures actual memory utilization in sequence models, enhancing understanding and optimization of model performance.
Researchers at UNC Chapel Hill introduced TACQ, a task-aware quantization method that preserves critical weight circuits, allowing large language models to maintain high accuracy even at ultra-low 2-bit precision compression.