FlashMLA: New standards for AI decoding on NVIDIA Hopper GPUs

DeepSeek FlashMLA

DeepSeek AI’s introduction of FlashMLA, an innovative decoding kernel technology for Multi-head Latent Attention (MLA), is a significant step in the continuous optimization of AI models. This open technology was developed specifically for the NVIDIA Hopper architecture and aims to dramatically improve the processing of variable sequence lengths in AI models.

Read more

AutoAgent Framework: How zero-code technologies accelerate the development of autonomous AI agents

AutoAgent Framework

The development of AI-powered autonomous systems is reaching a new level: with AutoAgent, a GitHub-based platform, customizable AI assistants are now made accessible without any programming knowledge. The combination of a fully automated process structure and an extremely resource-efficient design opens up new ways not only for companies but also for individuals to use personalized and powerful tools for their work.

Read more

Helix: The new vision-language-action technology for controlling humanoid robots

Figure AI Helix

The technological advancement of humanoid robots has reached a significant milestone. The company Figure AI presented Helix, a new vision-language-action (VLA) model platform for real-time control of humanoid robots via voice input. This innovation is characterized by the combination of visual and voice-based data processing as well as motion control and promises to fundamentally redefine the interaction between humans and machines.

Read more

Sakana AI: The AI CUDA Engineer is here!

Sakana AI

Sakana AI has opened a new chapter in the world of artificial intelligence and high-performance computing with the launch of AI CUDA Engineer. This innovative solution enables the automatic creation of highly optimized CUDA kernels for machine learning processes – a step that not only drastically increases the efficiency of models, but also shapes the future of sustainable AI systems.

Read more