Zum Hauptinhalt springen

I.AI Consulting & Training

Alibaba Unveils Trillion-Parameter AI Model Qwen3-Max

Alibaba has launched Qwen3-Max-Preview, a groundbreaking artificial intelligence model featuring over 1 trillion parameters that directly challenges OpenAI and Google’s dominance in the AI space. This massive model boasts a 262K context length and delivers exceptional performance across mathematical reasoning, programming, and scientific tasks.

The new GPT: Improvements without the breakthrough

OpenAI has released GPT-5 – a refined version of its AI model, now available via ChatGPT. A key new feature is the automatic switching between fast standard models and slower, more „thoughtful“ versions.

Cogito v2: Smarter Open-Source AI Models

Deep Cogito has introduced Cogito v2 – a new family of open-source AI models that can improve their own reasoning. The models range from 70 to 671 billion parameters, with the largest ranking among the most powerful open-source AIs in the world.

Mac app developed almost entirely with AI

Developer Indragie Karunaratne created the Mac app Context almost entirely using the AI tool Claude Code — writing only about 1,000 of the 20,000 lines of code himself. The app, built with Swift and SwiftUI, is designed for debugging MCP servers.

DeepSeek’s powerful, cheap AI update went unnoticed

DeepSeek’s latest R1 AI model, offering better performance at a lower cost, barely registered in the tech world. This is a stark contrast to early 2025, when the original R1 model sent tech stocks plummeting and raised concerns about the generative AI boom.

New Google App Runs AI Models Locally

Google has discreetly launched „Model Explorer,“ an app enabling users to download and run AI models directly on their devices. This move, unannounced by the tech giant, empowers developers and AI enthusiasts with greater control and privacy over their AI experiments.