Quanta AI Labs

50 / Dec 8 – Dec 14, 2025 · 28 RELEASES

AI News Week 50, 2025

Runway Gen-4.5, Kling 2.6/O1, and Mistral 3 lead video generation, with DeepSeek-V3.2 and Gemini Deep Think advancing reasoning.

Language Models (3)

  • Gemini 3 Deep Think

    Google · Reasoning

    New Deep Think mode for Google AI Ultra subscribers delivering meaningful improvements in reasoning capabilities.

  • Mistral 3

    Mistral AI · Text Generation

    Family of frontier open-source multimodal models including Mistral Large 3 and Ministral 3 with state-of-the-art performance.

  • DeepSeek-V3.2

    DeepSeek · Reasoning

    Reasoning-first models built for agents with world-leading reasoning, thinking in tool-use, and open source release.

Vision & Image (6)

  • Seedream 4.5

    ByteDance Seed · Text-to-Image

    Advanced image generation with reference image consistency, poster layout design, and accurate multi-image editing.

  • TUNA

    Multi-Institution · Text-to-Image

    Native unified multimodal model enabling image/video understanding and generation within a single framework.

  • LongCat-Image

    Meituan · Text-to-Image

    Advanced image understanding model with long-context capabilities and text-to-image generation.

  • Ovis-Image

    AIDC-AI · Text-to-Image

    7B text-to-image model optimized for high-quality text rendering under computational constraints.

  • PosterCopilot

    Research · Image Editing

    Layout reasoning and controllable editing for professional graphic design with multi-round refinement.

  • ThreadIt

    pravinl23 · Image Editing

    AI-enhanced clothing design app for sketching, enhancing, and launching garment designs directly to Shopify.

Video & Animation (6)

  • Runway Gen-4.5

    Runway · Text-to-Video

    New frontier for video generation with state-of-the-art motion quality, prompt adherence, and visual fidelity.

  • Kling 2.6 & O1

    Kuaishou · Text-to-Video

    World's first unified multimodal video model with integrated audio-visual generation, video editing, and Chain-of-Thought reasoning.

  • HunyuanVideo-1.5

    Tencent Hunyuan · Text-to-Video

    Leading lightweight video generation model with open-source training code, diffusers support, and community contributions.

  • PixVerse V5.5

    PixVerse · Text-to-Video

    Fast AI video generator with text-to-video, image-to-video, 1080p output, and optimized for social media content.

  • Live Avatar

    Alibaba Quark · Real-time Video

    Real-time streaming avatar generation at 20 FPS with 14B diffusion model supporting 10,000+ second infinite-length videos.

  • SteadyDancer

    NJU MCG · Video Editing

    Human image animation with first-frame preservation tackling spatio-temporal misalignment issues.

Audio & Speech (3)

  • VibeVoice Realtime

    Microsoft · Text-to-Speech

    Frontier open-source 0.5B text-to-speech model with real-time capabilities and responsible AI guidelines.

  • ViSAudio

    Research · Audio Processing

    End-to-end video-driven binaural spatial audio generation for immersive 3D sound creation.

  • Live2D Voice Assistant

    TEN AI · Voice Agents

    Tutorial for building interactive anime character voice assistants with real-time lip sync using TEN Framework.

3D & Spatial (1)

  • Lotus-2

    Research · Depth Estimation

    Advancing geometric dense prediction with powerful image generative models for depth and normal estimation.

Agents & Automation (4)

  • Advent of Agents

    Google Cloud · Agent Framework

    25 days of Zero to Production-Ready AI Agents tutorials on Google Cloud infrastructure.

  • Better Agents

    LangWatch · Agent Framework

    Standards and structure for building production-ready AI agents with proper documentation and patterns.

  • Agentic AI Foundation

    Linux Foundation · Agent Framework

    New foundation housing MCP, Goose, and AGENTS.md to advance open-source agentic AI development standards.

  • Reinforcement Trading Tutorial

    Ziad Francis · Agent Framework

    Educational repository for building reinforcement learning-based trading agents with Python.

Developer Tools (5)

  • HuggingFace Skills Training

    Hugging Face · Prompt Engineering

    Use Claude, Codex, or Gemini CLI to fine-tune open-source LLMs with SFT, DPO, and GRPO training methods.

  • Made With ML

    Goku Mohandas · AI Infrastructure

    Comprehensive course on designing, developing, deploying and iterating on production-grade ML applications.

  • Awesome MCP Servers

    punkpeye · AI Infrastructure

    Comprehensive collection of Model Context Protocol servers for browsers, databases, cloud platforms, and developer tools.

  • ADK ScrapeGraphAI

    Google · Web Scraping

    Third-party tool integration for Google Agent Development Kit enabling AI-powered web scraping and data extraction.

  • cuTile Python

    NVIDIA · AI Infrastructure

    Programming model for writing parallel GPU kernels in Python with simplified CUDA abstractions.

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.