India’s AI Challenger: Sarvam Unveils Indus — A New Chat App in a Crowded Field
India-based AI startup Sarvam has launched Indus, a new AI chat app available in beta on web, iOS, and Android, designed to compete with popular global services like ChatGPT, Google Gemini, and Anthropic Claude. Indus acts as the user interface for Sarvam’s recently introduced 105-billion-parameter large language model, letting people type or speak questions and receive responses in both text and audio. The app emphasises multilingual support tailored to India’s diverse languages and user needs, while also showcasing enterprise integrations and partnerships announced at the India AI Impact Summit.
Despite the buzz, Indus is still in early rollout with some limitations — for example, users currently cannot delete chat history without deleting their account, and access may be restricted initially due to limited compute capacity. Sarvam says it will expand availability over time and is seeking feedback from users as it refines the experience. The launch underscores India’s growing role as a key battleground in the global generative AI race.
Google’s Gemini 3.1 Pro Shatters Benchmarks in AI Reasoning Race
Google has unveiled Gemini 3.1 Pro, the latest upgrade to its flagship artificial intelligence model family, and it’s already turning heads with record-setting benchmark results. In independent tests — including the demanding reasoning benchmark ARC-AGI-2 — Gemini 3.1 Pro scored around 77 %, more than double the performance of its Gemini 3 Pro predecessor, highlighting a significant leap in multi-step reasoning, logic, and problem-solving capabilities that many competitors still struggle with.
Beyond pure numbers, the upgrade is being previewed across Google’s ecosystem — from the Gemini app and NotebookLM to enterprise and developer platforms like Vertex AI and Google AI Studio — giving a broad range of users earlier access to its improved reasoning and complex task handling. While rivals such as OpenAI and Anthropic continue to push their own models, Gemini 3.1 Pro’s early performance surge underscores how fierce and fast-moving the generative AI competition has become
Reddit Tests AI Shopping Search, Turning Chat into Commerce
Reddit is experimenting with a new AI-powered shopping search feature that blends its community-driven discussions with product discovery. In this pilot — currently rolling out to a small group of users in the United States — people searching for things like “best noise-canceling headphones” will now see interactive product carousels inside Reddit’s search results. These carousels show images, prices, and direct purchase links for items mentioned by users in relevant posts and comments, pulling catalogue data from Reddit’s shopping and advertising partners. The idea is to make product research faster and more useful by surfacing real community-recommended items rather than just discussion threads.
The move is part of Reddit’s larger push into e-commerce and monetisation, building on last year’s launch of Dynamic Product Ads. By linking its vast discussions directly to shoppable results, Reddit hopes to keep users engaged on the platform and potentially capture some of the product-search traffic that traditionally flows to Google and Amazon. The company says it will refine the shopping experience based on user feedback as the test continues.
Alibaba’s Qwen AI Shakes Up the Global AI Landscape
Alibaba’s open-source AI family, particularly the newly released Qwen 3.5, is making waves by directly challenging leading proprietary models like OpenAI’s GPT-5.2 and Anthropic Claude Opus 4.5. Qwen 3.5 uses a Mixture-of-Experts (MoE) architecture that activates only a fraction of its huge 397 billion parameters at a time, keeping performance high while lowering compute costs — a setup that makes it more economically efficient than many closed systems. The model also offers native multimodal reasoning and an expanded 1 million-token context window, enabling it to handle long documents, codebases, and multimodal inputs (e.g., text with images) more effectively.
Alibaba’s strategy underscores a broader shift in AI economics: by pushing open-source models with top-tier performance and broad language support, the company aims to make advanced AI accessible and affordable for enterprises worldwide. This approach could reshape cost dynamics in the industry, positioning open models as real alternatives to expensive proprietary systems and accelerating innovation and adoption.
YouTube Brings Conversational AI to Your TV Screen
Google’s video platform YouTube is testing a new conversational AI feature on smart TVs, gaming consoles, and streaming devices that lets viewers ask questions about the videos they’re watching without pausing or leaving the app. Users in the test group see an “Ask” button on their TV screen or can use their remote’s microphone to speak queries — for example, asking about recipe ingredients in a cooking video or artist info during music clips — and receive instant, contextual responses while the video keeps playing. The experiment extends YouTube’s existing mobile and web AI tools to the living room and supports multiple languages including English, Hindi, Spanish, Portuguese, and Korean.
This feature, powered by Google’s Gemini-based conversational AI, is currently limited to a small group of users (mostly Premium Labs testers aged 18+) and represents a broader push to make video watching more interactive. It reflects a trend where major platforms are blending AI chat tools into everyday media experiences, potentially reshaping how people engage with video content on large screens.