Records found: 5
'xAI released Grok-4-Fast, a single prompt-steerable model with a 2M-token window and tool-use RL that reduces token usage by about 40% while maintaining frontier benchmark performance.'
'New disclosures give rough per‑query energy estimates but leave major gaps about multimodal use, real efficiency gains, and whether demand will reach the scale companies expect'
'Microsoft's rStar2-Agent integrates code execution into the reasoning loop, allowing a 14B model to outperform larger systems on math benchmarks with shorter reasoning traces.'
Falcon-H1 from TII introduces a hybrid model combining attention and state space mechanisms, achieving performance on par with leading 70B parameter LLMs while optimizing efficiency and scalability.
NVIDIA researchers developed Dynamic Memory Sparsification (DMS), a novel method that compresses KV caches by 8× in Transformer-based LLMs, improving inference efficiency while maintaining accuracy.