The new AI reasoning champion challenging GPT-4.5

PLUS: AI detecting cancer with 99% accuracy, Claude's revolutionary 'think' feature & Rebuilding TikTok in America

In today’s agenda:

1️⃣ Tencent launches Hunyuan-T1 model outperforming GPT-4.5 across benchmarks

2️⃣ Anthropic's Claude 3.7 can now pause and rethink mid-task

3️⃣ Browser.ai raises $17M for smarter AI web navigation

Plus, some interesting news:

  • New AI detects endometrial cancer with 99.26% accuracy

  • Perplexity creates fact-first TikTok alternative

MAIN AI UPDATES / 24th Mar 2025

🤖 Hunyuan-T1 reasoning model 🤖
Tencent's new AI powerhouse

Tencent has launched Hunyuan-T1, a reasoning model designed to blend speed, accuracy, and efficiency. This model outperforms competitors like GPT-4.5 and o1 across multiple benchmarks with its optimized architecture for reasoning tasks. Hunyuan-T1 offers significant improvements in logical thinking and problem-solving capabilities, making it valuable for complex decision-making scenarios.

🧠 Claude 3.7 'think' tool 🧠
Real-time AI reasoning improvements

Anthropic has equipped Claude 3.7 with a revolutionary "think" tool that allows the AI to pause mid-process to reassess and refine its reasoning in real-time. Unlike extended thinking which analyzes before generating responses, this tool enables Claude to stop, reconsider, and adapt its approach during complex problem-solving scenarios. The enhancement particularly excels in high-stakes workflows requiring precise tool usage, with τ-Bench testing showing a 54% improvement in the airline domain pass metric.

🔍 Browser.ai raises $17M 🔍
Making AI agents web navigation easier

Browser.ai has secured $17 million in funding for its tool that helps AI agents navigate websites more effectively. The technology enables agents to better understand and interact with web content, potentially revolutionizing how automated systems operate online. This development represents a significant step forward in creating more capable AI assistants that can perform complex web-based tasks with greater autonomy and precision.

INTERESTING NEWS

🔬 ECgMLP Cancer Detection AI 🔬

Researchers from Charles Darwin University have developed ECgMLP, an AI model that identifies endometrial cancer with 99.26% accuracy from microscopic tissue images. This specialized model uses attention mechanisms to spot cancer cells that human specialists often miss, dramatically outperforming current diagnostic methods that only achieve 78-81% accuracy. The versatile system also excels at detecting colorectal, breast, and oral cancers with over 97% accuracy, promising to democratize access to expert-level cancer screening worldwide.

👀 Perplexity Vision 👀

Perplexity AI is developing a TikTok alternative that focuses on deep content discovery and truth-seeking powered by their advanced answer engine. Their approach aims to enhance the platform's utility while preserving its core function as a creative expression hub. The vision combines engaging social media elements with Perplexity's information retrieval capabilities to create a more informative experience. The concept is currently in planning stages, with details about implementation timeline and specific features yet to be announced. Early discussions suggest the platform could create a new category of social media that prioritizes factual content alongside entertainment.

📩 Have questions or feedback? Just reply to this email , we’d love to hear from you!

🔗 Stay connected: