Tag
117 articles
Google has released three new Gemini AI models but notably omitted the highly anticipated Gemini 3.5 Pro, prompting speculation about the company's strategic direction.
Google has released three new Gemini models—Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber—designed to be more cost-effective and token-efficient for agentic workloads.
Google has launched three new low-cost Gemini models to compete in the AI race, but the absence of a major rival to GPT-4o raises questions about its long-term strategy.
This article explains the technical concepts behind AI model scaling, token efficiency, and the strategic trade-offs between performance and deployment efficiency in modern AI systems.
The release of China's largest open-weight AI model, Kimi K3, has reignited policy debates in Washington over the risks and regulatory responses to such technologies.
This article explains Mixture of Experts (MoE) AI models, how they work like teams of specialists, and why they're important for efficient AI performance.
Google has updated its Gemini AI usage tracking system, introducing a new rating method that could reduce the number of responses users receive. The change affects both free and paid users, requiring more careful monitoring of AI interactions.
This article explains what AI models are, how Kimi K3's success has sparked concern in the U.S., and why the global AI race matters.
NVIDIA has released Nemotron 3 Embed, an open embedding collection featuring three models, with the 8B checkpoint ranking #1 on the RTEB benchmark.
Thinking Machines has launched its first open AI model, Inkling, marking a strategic shift from generic AI approaches toward specialized solutions. The move signals a growing industry trend toward more customizable AI infrastructure.
Mistral AI introduces Robostral Navigate, an 8B model that enables robots to navigate complex environments using only a single RGB camera, without requiring LiDAR or depth sensors.
The AI industry is shifting away from the belief that the biggest models win, with companies now prioritizing cost, task-specific performance, and control over sheer model size.
