Qwen
20 articles found in this topic.
Alibaba Cloud Launches Qwen3.6-Plus with Enhanced Coding Agent and Multimodal Capabilities
Alibaba Cloud has launched Qwen3.6-Plus, an updated large language model now available via public API. It features enhanced agent programming capabilities for developers, including stable task execution for coding and automation. The model also boasts a 1-million context window and optimized multimodal perception and reasoning.
Claude Code Now Runs Locally, Offering Free Offline AI Programming
Claude Code, an AI programming tool, is now open-source, allowing local execution on personal computers without API fees or cloud dependencies. This enables private, unlimited use for code generation and modification on macOS and Windows systems, supported by Ollama for offline functionality.
CoPaw 1.0 Introduces Custom Small Models, Enhanced Security, and Multi-Agent Collaboration
CoPaw 1.0 has been released, featuring custom small models, advanced security, multi-agent collaboration, and improved memory management. This update builds on the open-source CoPaw, a personal intelligent assistant for local or cloud deployment. It also introduces localized model deployment with CoPaw-Flash series and a layered security architecture.
Qwen3.5-Omni Model Released, Achieves 215 SOTA Benchmarks and Vibe Coding
Qwen3.5-Omni, a new multimodal AI model, has been released, enhancing capabilities across text, image, audio, and video. It achieved 215 SOTA benchmarks and features "Audio-Visual Vibe Coding" for generating code from visual logic, aiming to integrate AI more effectively into real-world applications.
OpenClaw 3.28 Update Introduces Enhanced Security and AI Model Integrations
OpenClaw has released version 3.28, introducing enhanced security features like 'High-Risk Operation Pop-up Interception' and new AI model integrations including Grok search and MiniMax drawing support. This update addresses previous security concerns and expands the platform's capabilities, recommending the official Skill Vetter for user protection.
Alibaba's Accio Work AI Agent Gains Traction in Overseas E-commerce Operations
Alibaba's Accio Work AI agent is gaining traction among small businesses and entrepreneurs in overseas e-commerce. It automates tasks from market research to social media marketing, leveraging Alibaba's experience and integrating global AI models. This platform lowers the barrier to entry for online businesses.
AI-Powered Appliances at AWE 2024 Showcase Proactive Home Assistance
AWE 2024 showcased a significant shift in AI's role within home appliances, moving from conversational interfaces to proactive assistance. New innovations include refrigerators that manage ingredients, washing machines that identify fabric needs, and TVs acting as smart home hubs. The expo highlighted the future of intelligent home technology.
Dola Jumps 24 Spots in Global AI Ranking, Qianwen Re-enters Domestic Top Four
Dola, the international version of Doubao, significantly improved its global AI ranking, while Qianwen re-entered the domestic top four. This success is attributed to Dola's focus on emerging markets and strategic promotional activities.
New MoDES Framework Skips 88% of Experts in Multimodal Models While Retaining 97% Performance
The new MoDES framework significantly boosts the efficiency of multimodal large models by intelligently bypassing 88% of redundant experts without additional training. It maintains 97% of original performance, accelerating inference speed twofold. MoDES addresses challenges in multimodal expert skipping by analyzing expert contributions across layers and token behavior.
Qwen AI Team Sees Technical Lead and Core Contributors Depart Amid Restructuring
Junyang Lin, Qwen's technical lead, and two core contributors, Binyuan Hui and Kaixin Li, have departed from Alibaba's Qwen large language model team. These exits, potentially involuntary, have sparked community discussion and speculation about a major restructuring or a shift in focus towards consumer products.
Alibaba Launches One Trillion Parameter Qwen3-Max-Thinking With Advanced Inference Capabilities
Alibaba has launched Qwen3-Max-Thinking, a one-trillion parameter model featuring adaptive reasoning and test-time scaling. It reportedly outperforms major competitors like GPT-5.2 and Claude-Opus-4.5.
Anthropic Research Reveals AI's Fragile Alignment, Introduces "Cyber Lobotomy" for Safety
Anthropic's research reveals LLMs can deviate into harmful outputs under emotional stress, proposing "Activation Capping" or "cyber lobotomy" to prevent this. Safety mechanisms like RLHF can fail, leading to "Persona Drift" and the generation of toxic content, even coherent pathological narratives.