Latest
OpenAI has released GPT-6 Sol and GPT-6 Luna, cutting API prices by 50% and introducing a 90% discount on cached input. Here is what the numbers mean for per-task costs, the Opus 5.5 launch the same day, and how to choose models for AI agents.
A look at Xiaomi's MiMo-V2.6 mixed RL training, the Flash model that nears top benchmark scores at roughly a third of Pro's API price, and the 7,000+ published RL training environments.
At least 45 US data center projects, valued at about $68 billion, were blocked or delayed by local opposition in April–June 2026. Here is what the figure does and does not mean, and how New York's voluntary guidance frames who bears local costs.
Anthropic has announced its new flagship model, Claude Opus 5.5, which matches the reasoning performance of Claude Fable 5.1 while cutting costs by 40% compared to Opus 5. The model also takes the top spot on the Intelligence Index and features a refreshed 'Claudish' writing style.
TrendForce data shows the global DRAM module market surged 59% to $21.2 billion in 2025, but leader Kingston grew only 48%—below the market average—and saw its share slide from 66% to 62%. Here's what happened at No. 3, the only rank that actually moved.
A closer look at MediaTek's new Googlebook SoC compared with the older Kompanio chip reveals a shared core design with selective upgrades—and lingering questions about an unannounced Lenovo device.
Googlebook has opened preorders as a premium PC lineup from five manufacturers, five models, starting at $899. Here's how it layers the web, apps, and an isolated Linux environment on an Android foundation—and the uncertainties still lingering around pricing and app quality.
A widely shared estimate says 100 million Muse users would need about 1.58 million CPUs and 800PB of RAM. We correct the math and separate vCPU allocation, idle memory and storage to see what infrastructure demand might really look like.
Soochow University and LONGi placed insulating zirconia particles at a buried interface to cut charge losses in a tandem solar cell. Here is how it works, what the certified figure means, and what the 2,000-hour test shows.
AWS has clarified Strands Agents as an open-source harness SDK. Drawing on primary sources, this piece examines its continuity with the 2025 code release, how responsibilities are divided from AgentCore Harness, and what "any model, any cloud" actually means in practice.
