KI-Nachrichten Hub

AI Coding Agents

Cursor 2.0 Introduces Background AI Agents for Autonomous Coding

Cursor releases version 2.0 with background AI agents that can autonomously work on coding tasks while developers focus on other work. The feature supports parallel task execution and integrates with CI/CD pipelines.

cursor coding agents +2

Feb 17, 2026 · Cursor

Lesen

Text Models

Gemini 3.1 Ultra Launches with 2M Token Context Window

Google DeepMind releases Gemini 3.1 Ultra with an industry-leading 2 million token context window and improved multimodal understanding. The model processes entire codebases, lengthy documents, and multi-hour video inputs natively.

google gemini context-window +2

Feb 16, 2026 · Google

Lesen

Video Models

Runway Gen-4 Turbo Reduces Generation Time by 80%

Runway launches Gen-4 Turbo, a high-speed variant of its video generation model that produces results 80% faster than Gen-3 Alpha. The update enables near-real-time video iteration for professional creative workflows.

runway gen-4 video-generation +2

Feb 16, 2026 · Runway

Lesen

AI Agents & Tools

Google Launches AI Agent Builder Platform for Enterprises

Google Cloud introduces AI Agent Builder, a low-code platform enabling enterprises to create, deploy, and manage custom AI agents. The platform supports multi-agent orchestration and integrates with Google Workspace and third-party tools.

google agent-builder enterprise +2

Feb 15, 2026 · Google

Lesen

Image Models

Stable Diffusion 4.0 Released as Fully Open Source

Stability AI releases Stable Diffusion 4.0 under a fully open-source license, featuring a new transformer-based architecture that rivals closed-source models in quality. The release includes training code, data pipelines, and model weights.

stability-ai stable-diffusion open-source +2

Feb 15, 2026 · Stability AI

Lesen

AI Coding Agents

Windsurf Cascade Achieves 94% Codebase Awareness Score

Windsurf announces that its Cascade AI engine has reached a 94% codebase awareness score on the new CodeContext benchmark, setting an industry record for understanding project architecture and cross-file dependencies.

windsurf cascade coding +2

Feb 14, 2026 · Windsurf

Lesen

Open Source AI

Qwen 3 MoE Leads Open Source Coding Benchmarks

Alibaba Cloud releases Qwen 3 MoE, which takes the top position on open-source coding benchmarks including HumanEval, MBPP, and LiveCodeBench. The model features a 128K context window and native tool-use capabilities.

alibaba qwen coding +2

Feb 14, 2026 · Alibaba

Lesen

Text Models

Grok 4.1 Tops LM Arena Leaderboard Across All Categories

xAI releases Grok 4.1, which achieves the top position on the LM Arena leaderboard across all evaluation categories including reasoning, coding, creative writing, and mathematical problem-solving.

xai grok leaderboard +2

Feb 12, 2026 · xAI

Lesen

Video Models

Kling 2.0 Introduces Multi-Shot Narrative Video Creation

Kuaishou Technology releases Kling 2.0 with multi-shot narrative video generation, enabling users to create coherent story-driven videos with consistent characters across multiple scenes and camera angles.

kling kuaishou video-generation +2

Feb 12, 2026 · Kuaishou

Lesen

Image Models

DALL-E 4 Adds Video Frame Generation and Animation Support

OpenAI unveils DALL-E 4 with the ability to generate animated sequences and video keyframes from text prompts. The model bridges the gap between still image generation and full video creation with smooth interpolation.

openai dall-e animation +2

Feb 11, 2026 · OpenAI

Lesen

AI Agents & Tools

Microsoft Copilot Studio Adds Custom Agent Development Kit

Microsoft releases a custom agent development kit for Copilot Studio, enabling developers to build AI agents with enterprise connectors, memory systems, and governance controls. The SDK supports TypeScript and Python.

microsoft copilot-studio agent-sdk +2

Feb 11, 2026 · Microsoft

Lesen

AI Coding Agents

Devin 2.0 Sets New SWE-Bench Record at 71.4%

Cognition Labs releases Devin 2.0, achieving a 71.4% score on the SWE-Bench Verified benchmark. The autonomous coding agent can now handle multi-step debugging, deploy applications, and collaborate with human developers in real-time.

devin autonomous swe-bench +2

Feb 10, 2026 · Cognition

Lesen