Quanta AI Labs

05 / Jan 26 – Feb 1, 2026 · 25 RELEASES

AI News Week 5, 2026

DeepSeek-OCR-2 and Gemini 3 Flash advance vision, Qwen3-Max pushes reasoning, with HunyuanImage 3.0 and Project Genie leading generation.

Language Models (3)

  • Qwen3-Max Thinking

    Qwen · Reasoning

    Qwen3-Max with enhanced thinking capabilities, pushing reasoning boundaries in open models.

  • Kimi-K2.5

    Moonshot AI · Reasoning

    Powerful reasoning model with native INT4 quantization and coding agent framework.

  • SERA-8B

    AllenAI · Code Generation

    Specialized 8B model achieving high performance on SWE-bench Verified with 32k context.

Vision & Image (3)

  • DeepSeek-OCR-2

    DeepSeek AI · OCR & Document

    Visual Causal Flow model for OCR with improved structured data extraction and reasoning.

  • HunyuanImage 3.0

    Tencent · Text-to-Image

    Powerful native multimodal model for image generation with instruction reasoning and editing.

  • Qwen Video Turbo

    Wuli-art · Text-to-Image

    Turbo LoRA for Qwen-Image-2512 achieving high quality generation in just 2 steps.

Video & Animation (3)

  • Lucy 2

    Decart AI · Video Editing

    Real-time world transformation model enabling live video editing and physics simulation.

  • TeleStyle

    Tele-AI · Video Editing

    Content-preserving style transfer framework for images and videos using Diffusion Transformers.

  • Luma Ray3

    Luma AI · Text-to-Video

    New model for hybrid-AI workflows in video generation and acting performances.

Audio & Speech (3)

  • MirrorMate

    OrangeKame3 · Voice Agents

    Self-hosted personalized AI smart mirror with voice interaction and avatar memory.

  • Qwen3-ASR 1.7B

    Qwen · Speech-to-Text

    Efficient 1.7B parameter automatic speech recognition model with high accuracy.

  • MiniMax Music 2.5

    MiniMax · Music Generation

    New music generation model with improved composition and high-fidelity audio.

3D & Spatial (1)

  • Project Genie

    Google DeepMind · 3D Generation

    AI world model for generating infinite, interactive playable worlds from images.

Agents & Automation (8)

  • Gemini 3 Flash Agentic Vision

    Google · Agent Framework

    New agentic vision capability converting image understanding into a dynamic, interactive process.

  • Claude Code Tools

    Community · Agent Framework

    Comprehensive collection of agents, skills, and configuration for Claude Code CLI.

  • OpenClaw

    OpenClaw · Agent Framework

    Personal AI assistant that runs locally with full system access and persistent memory.

  • Moltbook

    Moltbook · Agent Orchestration

    Social network built exclusively for AI agents to share, discuss, and interact.

  • Crush

    Charmbracelet · Coding Agents

    Glamorous agentic coding tool for terminal with MCP support and local models.

  • LingBot World

    RobbyAnt · Robotics

    Embodied intelligence platform for robotics with spatial perception and world models.

  • SJinn

    SJinn · Agent Orchestration

    Professional AI agent for image, video, audio, and 3D content creation workflows.

  • Mosi MOVA

    Mosi · Agent Framework

    Next-generation interactive AI building context-aware foundation models.

Developer Tools (3)

  • DevTools

    DevTools · AI Infrastructure

    Local-first API testing and flow automation tool that records browser traffic to YAML.

  • Semantic Highlight

    Zilliz · Inference Engine

    Lightweight bilingual semantic highlight model for search and retrieval optimization.

  • Earth-2 Open Models

    NVIDIA · AI Infrastructure

    Fully open family of accelerated AI weather models for global forecasting.

Research & Safety (1)

  • System Prompts Leaks

    Community · Alignment & Safety

    Collection of extracted system prompts from top chatbots like ChatGPT and Claude.

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