AI
RecursiveMAS Cuts Multi-Agent Token Use 75%
Researchers at UIUC and Stanford have developed RecursiveMAS, a multi-agent framework that routes inter-agent communication through…
Tag: inference efficiency
AI
AI Architecture in 2026: Efficiency Over Scale
Five converging developments in mid-2026 — RecursiveMAS's 2.4× inference speedup, OpenAI's Parameter Golf challenge, 5% enterprise…
NVIDIA
ZAYA1-8B Achieves GPT-5 Performance with 760M Parameters
Zyphra's ZAYA1-8B achieves GPT-5 performance with just 760M active parameters, while Subquadratic claims 1,000x efficiency gains…
Google
AI Efficiency Gains Heat Up
Miami startup Subquadratic claims 1,000x AI efficiency gains through subquadratic architecture while Google delivers practical 3x…
AI Agents
Nemotron 3 Nano Omni Delivers 9x Efficiency for AI Agents
NVIDIA's Nemotron 3 Nano Omni unifies vision, audio, and language in a single model, delivering 9x…
Microsoft
AI Architecture Advances: New Transformer Models Cut Inference Costs 41%
Microsoft's new MAI-Image-2-Efficient model delivers 41% cost reduction and 22% faster inference, while NVIDIA emphasizes cost-per-token…
AI
AI Architecture Advances Drive Efficiency Gains Across Models
Major AI companies are launching architecturally optimized models that dramatically reduce inference costs while improving performance.…
Enterprise
AI Architecture Advances Drive 40% Cost Reductions in 2025
Major AI architecture breakthroughs in 2025 are delivering 40% cost reductions through optimized transformer designs, parameter-efficient…