Track global AI model launches, company moves, open-source projects, and infrastructure updates worth paying attention to.
Liquid AI LFM2.5 is a 1.2B lightweight model for edge devices. Runs on phones, supports multimodal, with Ollama & vLLM deployment guide.
GLM-4.7 just launched with stronger agentic coding, tool calling and reasoning. This quick guide covers transformers inference, vLLM/SGLang local serving (OpenAI-style API), plus Thinking/Preserved Thinking switches and key params.
GPT-5.2 is OpenAI’s newest model family for professional work and long-running agents, improving long-context reasoning, tool use, vision, and coding, plus updated API names and pricing.
US Commerce Department to approve Nvidia H200 AI chip exports to China, signaling easing of tech restrictions amid stabilizing US-China relations.
DeepSeek-V3.2 is the latest open-source LLM rivaling GPT-5 and Gemini-3.0-Pro, featuring DSA sparse attention and gold-medal results in IMO/IOI 2025.
Qwen3-Omni-modal model for text, image, audio and video with real-time speech. Thinker–Talker + MoE, multi-codebook for low latency; 119 languages; vLLM/Transformers tips.
Qwen3-Next series: a hybrid architecture with Gated DeltaNet × Gated Attention. 80B total parameters with ~3B active per step, optimized for long context, high concurrency, and low latency. Instruct and Thinking target production chat and deep reasoning respectively.
DeepSeek V3.1 adds Think/Non‑Think hybrid reasoning, stronger coding and Agent abilities, 128K context and MoE efficiency, plus Anthropic API support for Claude Code.
Immersive Translate extension security vulnerability exposes massive user data leakage through webpage snapshot feature. Analysis of technical flaws, community backlash, and official response.
GLM-4.5 is Zhipu AI's new agent-native model, featuring efficient MoE architecture, strong reasoning, coding, and open-source ecosystem.
