AI Platform Hits 1.33 Trillion Daily Token Throughput Milestone
An AI infrastructure platform reports record daily token throughput of 1.33 trillion and over 10.86 million daily API calls, with explosive growth since August…
Llama Mac App Adds Simple Request Builder for llama.cpp REST API
The Llama app for Mac now includes a simple request builder for llama.cpp's REST API, making it easier to test and interact with local LLM endpoints.
MicroDuck Robot Learns Backflip, Trained on a MacBook Pro
A developer shows a small robot named MicroDuck performing a clean backflip, trained entirely on a MacBook Pro, with the repo to be open sourced soon.
Tencent's Hy4 Preview: Open-Source 770B MoE Model Lifts Inference Throughput by 31.8%
Tencent's Hy4 preview, a 770B-parameter open-source MoE model with 1M context, lifts end-to-end throughput by 31.8% via operator fusion and communication optimi…
Tencent's Hy4 Preview: A Free, Open-Source Coding Model That Impresses
Tencent's Hy4 preview model is free, open source, and nearly overlooked, yet it delivers impressive coding results—like generating a 3D isometric city with cars…
Skyfire: Building the Intelligent Global Settlement Layer for the AI Economy
Skyfire is building an intelligent global settlement layer for the AI economy, positioned above every model and beneath every agent, to enable autonomous paymen…
Miora × MiniMax H3 Launch 'Bestiary' Global Short Film Contest for Strange Creatures
Miora and MiniMax H3 announce Bestiary, a global short film contest inviting creators to submit strange-creature narrative shorts, with $8,000 in cash prizes an…
Requesty Partners with Runware to Offer GLM 5.3 Flash and DeepSeek v4 Flash Inference Endpoints
Requesty has partnered with Runware to bring new inference endpoints to its platform, including GLM 5.3 Flash and DeepSeek v4 Flash 0731, now available for user…
WinkLink Pitches Itself as the Settlement Layer for the AI Economy
WinkLink, an oracle project in the TRON ecosystem, says it is building an intelligent global settlement layer for the AI economy, positioning itself between AI…
Sliding Window Attention: A Simple Post-Training Switch That Beats Linear Attention
A new result shows that switching to sliding-window attention with attention sinks can outperform linear attention post-training at no extra cost. Here's what S…