- Projects of AI
- Posts
- Google Gemini 2.5 Pro tops benchmarks with built-in reasoning.
Google Gemini 2.5 Pro tops benchmarks with built-in reasoning.
Plus: Microsoft's AI security agents.
Amandeep Singh
March 26, 2025
Together with
Welcome Humans🤖,
Here is what we have Today:
🧠 Gemini 2.5 Pro tops benchmarks, making other AIs look like kindergarteners.
📸 Alibaba turns your photo into a movable 3D avatar.
🕵️♀️ Microsoft unleashes 11 AI agents to fight digital crime.
💻 DeepSeek's massive model fits on your high-end computer.
💼 New Job Opportunities
Google just dropped Gemini 2.5 Pro, its most advanced AI yet—packed with built-in reasoning and topping key benchmarks.
Image Source: Google
The Breakdown:
🚀 Smarter Reasoning, Smarter Answers
Unlike traditional models, Gemini 2.5 reasons step by step, refining its solutions before responding—just like a human problem solver. That means sharper accuracy and deeper context understanding.
🎨 Multimodal Monster
With a 1 million token context window (soon expanding to 2 million!), it seamlessly processes text, audio, images, video, and full codebases.
💻 Coding on Another Level
Gemini 2.5 Pro smashed the SWE-Bench Verified benchmark at 63.8%, generating everything from complex web apps to interactive, executable projects—straight from a single prompt.
📊 Top of the Charts
It outperforms rivals like GPT-4.5 and Grok 3 in math and science, scoring a record-breaking 18.8% on Humanity’s Last Exam—a test designed to push the boundaries of human knowledge.
🌎 Where to Try It
Gemini 2.5 Pro is now live in Google AI Studio and the Gemini app for Gemini Advanced users, with Vertex AI integration coming soon.
⚡ Why It Matters
Google is making AI "think" by default—no premium required. Despite OpenAI stealing the spotlight, Gemini 2.5 Pro proves Google is still in the game. But with GPT-5 and rivals lurking, how long will the crown last?
Together with Superhuman AI
Find out why 1M+ professionals read Superhuman AI daily.
In 2 years you will be working for AI
Or an AI will be working for you
Here's how you can future-proof yourself:
Join the Superhuman AI newsletter – read by 1M+ people at top companies
Master AI tools, tutorials, and news in just 3 minutes a day
Become 10X more productive using AI
Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.
DeepSeek
China’s AI powerhouse just pulled off another surprise. DeepSeek quietly rolled out V3-0324—an advanced 641GB model that runs on high-end personal computers and comes with a game-changing MIT open-source license.
Image Source: DeepSeek
The details:
Smarter computing: Uses a Mixture-of-Experts architecture, activating just 37B parameters per token—slashing hardware demands.
Mac-friendly AI: Testers confirm it runs smoothly on Apple’s Mac Studio, making top-tier AI accessible beyond data centers.
Sharper skills: Early users report major improvements in math and coding—some call it the best non-reasoning model available.
Truly open-source: Unlike its predecessor, V3-0324 is released under the highly permissive MIT License.
Why it matters:
DeepSeek isn’t slowing down. What seems like a minor upgrade is making major waves, fueling speculation about the upcoming R2 release. Could DeepSeek be positioning itself as the new AI leader? Stay tuned.
Trending Today
Japan has unveiled its first Hydrogen Coca-Cola vending machine. The new vending machines are more environmentally friendly and has a generator module on the side, that's where the Hydrogen Cartridges are installed generating power to operate making the smaller carbon footprint.
Alibaba has unveiled the Large Animatable Human Reconstruction Model (LHM) that transforms a single photo into a lifelike, movable 3D human avatar enabling the creation of realistic digital humans with diverse poses and clothing styles, opening new possibilities in movies, video games, and online shopping. The model is available as open-source code on GitHub and HuggingFace.
Microsoft has introduced 11 new AI agents for its Security Copilot platform to automate repetitive cybersecurity tasks and enhance efficiency. These agents handle jobs like triaging phishing emails and generating regulatory notifications after data breaches, easing the burden on security teams managing thousands of alerts daily.
Alibaba has introduced Qwen2.5-VL-32B, a powerful multimodal AI model with strong visual and text processing capabilities. It excels in mathematical reasoning, detailed image understanding, and aligns well with human conversational preferences. It outperforms larger models like Mistral-Small-3.1-24B and Gemma-3-27B-IT in several benchmarks.
Recommended Reading
If WE had to recommend other newsletters
AI Tool Report is an Newsletter, we enjoy reading daily. AI Tool Report delivers top techniques on how AI can transform your business filled with practical tips, real-world examples.
A Smart Bear is for people who love thinking about strategy, startups, product, marketing, decision-making, and founder psychology.
Job Opportunities
AI News
Source: Ideogram
Ideas? Comments? Complaints?
We read your emails, comments and poll replies daily.
Get the most important AI, tech, and science news in a free daily email.
New Here? Subscribe!
Sponsorship Slots Open for March and Reach over 800+ active readers. (Now 40% off) 🤯
What`d you think of today`s edition? |
Until next time, Stay Informed!
Reply