Amazon Alexa+ with AI Assistant

Amazon introduces Alexa+, an AI-powered personal assistant that is smarter, more conversational, and action-oriented. Available for free to Prime members ($19.99/month otherwise), Alexa+ enhances home automation, shopping, organization, and entertainment.

Key Features:

• Conversational AI: Understands informal speech and complex queries.

• Task Automation: Books reservations, orders groceries, tracks deliveries, and manages smart homes.

• Agentic Capabilities: Independently completes online tasks (e.g., finding and scheduling repairs).

• Personalization: Remembers preferences, past purchases, and important user details.

• Cross-Device Experience: Available on Echo devices, mobile apps, and browsers, with context-aware continuity.

• Deep Knowledge: Uses LLMs and user-provided data (documents, emails, images) for insights and actions.

• Privacy & Security: Built with AWS security, centralized privacy controls, and a dedicated privacy dashboard.

Alexa+ is positioned as an AI-powered lifestyle assistant, integrating deeply into users' daily lives while maintaining ease of use.

Link to details

Nvidia Evo 2 AI Model For Biomolecular Sciences

Evo 2 is a powerful new AI model for genomics, developed by Arc Institute and Stanford University, and is now publicly available. Built on NVIDIA DGX Cloud and accessible via the NVIDIA BioNeMo platform, Evo 2 was trained on nearly 9 trillion nucleotides and can predict protein structures, identify novel molecules, and assess gene mutations.

The model enables faster, more precise biological design, transforming research in healthcare, agriculture, and materials science. It can analyze long genetic sequences (up to 1 million tokens), helping researchers link genetic variations to diseases like breast cancer with high accuracy. Evo 2 also has applications in climate-resilient crops and bioengineering.

NVIDIA contributed computing power via 2,000 H100 GPUs and AI expertise, accelerating Evo 2’s capabilities. The project is part of Arc Institute’s broader mission to support long-term scientific research in areas such as cancer, immune dysfunction, and neurodegeneration.

Link to the AI Tool

Microsoft Open-Sources Phi 4 Mini & Multimodal

Microsoft has released two new small language models: Phi-4-multimodal and Phi-4-mini.

• Phi-4-multimodal (5.6B parameters):

  • Microsoft’s first multimodal model, integrating speech, vision, and text in a unified architecture.

  • Uses a Mixture-of-LoRAs approach for efficient processing.

  • Optimized for on-device and edge computing with low latency.

  • Outperforms existing multimodal models like Gemini-2.0-Flash and InternOmni in speech recognition, speech translation, OCR, and vision-language tasks.

• Phi-4-mini (3.8B parameters):

  • Dense, decoder-only transformer with grouped-query attention.

  • Supports 200,000 vocabulary tokens and sequences up to 128,000 tokens.

  • Strong performance in math, coding, reasoning, and instruction-following, matching or outperforming larger models.

Both models are open-source (MIT-licensed) and available on Azure AI Foundry, HuggingFace, and the NVIDIA API Catalog. Phi-4 models excel in efficiency, reasoning, and scalability, making them ideal for AI applications in low-resource environments, real-time inference, and industry use cases like healthcare, manufacturing, and finance.

Link to technical report

Perplexity's New Voice Mode

Imagine having a real-time conversation with AI that not only understands multiple languages but also pulls in the latest information as you speak. That’s exactly what Perplexity Voice Mode delivers, the only experience that seamlessly combines real-time voice and real-time knowledge across languages.

🔹 Available now on iOS (Android coming soon!)
🔹 Just update the app and start talking, no more typing, no more sifting through links.😂

Voice + Real-Time Information is the ultimate disruption to Search. Once you experience an interactive, knowledge-rich conversation, going back to keyword searches and scrolling through pages will feel like trading a sports car for a horse. 🏎️➡️🐎

Link to launch

Introducing OpenResearcherAI.com