Top AI News Mistral, OpenAI, LG, Gemma, 3D Model and more!



AI Summary

Weekly AI Model Updates

  • Gemma 3: Most capable model for a single GPU, 27 billion parameters, supports 140 languages, multimodal with 128,000 token context.
  • Mix Small 3.1: Multimodal, multilingual, 128,000 token context, high performance, Apache 2.0 license.
  • XA1 from LG: Deep reasoning model for math, science, coding; versions at 32B, 7.8B, and 2.4B.
  • Hunan 3D Generation: Improved versions MV and mini, open source with superior performance.
  • Hunan T1: Hybrid transformer architecture with strong logic, longer context, and excellent long text processing.
  • Ernie 4.5 & X1: Deep thinking multimodal model, outperforms GPT 4.5 at reduced cost.
  • Command A: High speed, efficiency, optimized for minimal compute; supports 256,000 context length.
  • Gemini 2.0: Native image generation and conversational editing for images.
  • Cosmos Transfer One: Conditional world generation with multimodal support, creating photorealistic images.
  • Roblox 3D Generation: Create objects directly in Roblox by typing.
  • Stability AI’s Multi-View Video Generation: Transform 2D images into 3D videos with camera control.
  • Open Maners AI: Automation tool for complex tasks.
  • Fish CAD: Generates 3D models from text input.
  • Anthropic Code Web Search: Performs efficient web searches.
  • OpenAI’s Speech-to-Text Models: New models outperform previous versions.
  • Image Editing: Easily modify images with specified changes.
  • Deeper Search: Enhanced reasoning and search capabilities.
  • Mirage: Create custom virtual ads with virtual spokespersons.
  • Mind Maps in Note LM: Automatically generates mind maps from uploaded content.
  • Claw 3.7 and Wave 5: Advanced modules for enhanced modeling and prediction.
  • Jama Package: Fine-tune models for multimodality and tokenization with simple commands.