Navigate Select ESC Close

Gemini 3 Rumors Are CONFIRMED, It's VERY GOOD

2025-11-18 Science & Technology
16.8k
1.2k
100
Matt Wolfe
Matt Wolfe
956.0k subscribers

Google Gemini 3 Launch: Deep Dive into New Flagship Model and Agent Mode Capabilities

Experience the first unfiltered look at Google Gemini 3, analyzing its benchmark dominance, enhanced reasoning, and powerful new web-browsing Agent Mode.

Short Summary

  • Gemini 3 achieved new state-of-the-art results on expert benchmarks like Humanity’s Last Exam and GPQA Diamond.
  • The model shows massive improvements in reasoning, coding, multimodality, and long context handling.
  • The new Agent Mode can read emails, browse the web, and perform complex, multi-step tasks like booking reservations.
  • Deepthink mode offers enhanced reasoning for the hardest problems but is currently limited access.

This briefing dissects the core features of Gemini 3, moving beyond marketing claims to test its performance on complex logic puzzles, code generation (including fully functional 3D games), and automated workflows via Agent Mode. Assess whether this release justifies dethroning current leading LLMs based on empirical demonstrations.

Unlock all features

FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.

Description

#GeminiAI #AINews #Google The new Google Gemini 3 model is finally here...and yes, I got early access! Google said, “Matt, don’t share this with the world yet”… so I waited until the exact moment Gemini 3 went public to drop this deep-dive. In this video, I break down everything Gemini 3 can actually do, including its new agent mode, speed improvements, reasoning upgrades, and how it stacks up against other models like GPT-5, Claude, and Grok. You’ll see the good, the bad, and the parts Google probably didn’t expect me to show. I also test Gemini 3 across real workflows, benchmarks, and creative tasks — including Humanity’s Last Exam and GPQA Diamond, so you can see where it shines and where it still struggles. If you want the honest, unfiltered look at whether Gemini 3 can dethrone ChatGPT… this is it. Right now, Make is giving my audience one month free of the Pro Plan when you sign up using this link: https://www.make.com/en/register?promo=mattwolfe&utm_source=mattwolfe&utm_medium=influencer&utm_campaign=mattwolfe-integration-nov25 Discover More: 🛠️ Explore AI Tools & News: https://futuretools.io/ 📰 Weekly Newsletter: https://futuretools.io/newsletter 🎙️ The Next Wave Podcast: https://youtube.com/@TheNextWavePod Socials: ❌ Twiter/X: https://x.com/mreflow 🖼️ Instagram: https://instagram.com/mr.eflow 🧵 Threads: https://www.threads.net/@mr.eflow 🟦 LinkedIn: https://www.linkedin.com/in/matt-wolfe-30841712/ 👍 Facebook: https://www.facebook.com/mattrwolfe Resources From Today's Video: Let’s work together! - Brand, sponsorship & business inquiries: [email protected] #AINews #AITools #ArtificialIntelligence Time Stamps: 00:00 - Introduction & Early Access 00:19 - What is Gemini 3? Overview & Capabilities 01:05 - What's NEW & Benchmarks 02:53 - Availability & Where to Use Gemini 3 03:51 - Deep Think & Gemini Agent Announcements 04:15 - Agent Mode: Features & Cautions 04:57 - Productivity with AI & Make.com Sponsor 05:20 - Building an Audience Growth Agent 05:55 - Make.com Features & Offer 07:02 - Testing Gemini 3 in Google AI Studio 08:20 - Logic & Probability Puzzle Test 09:32 - Research Paper Summarization & Visualization 10:07 - "Attention is All You Need" Visualization 12:30 - Turning Notes into a YouTube Narrative 13:25 - HTML/CSS Animation Demo 14:10 - Minecraft-like Game Generation 16:11 - Advanced Game Generation (No Constraints) 17:15 - Turn-Based Strategy Game Demo 19:07 - Vampire Survivors Clone & Game Feedback 20:24 - Music Generation: Happy Birthday & Original Song 22:32 - Agent Mode: Research & News Rundown 24:34 - Executive Assistant Demo 26:16- Booking a Restaurant with Agent Mode 28:20 - Final Thoughts & Impressions

Top Comments (10)

@iks8622 2025-11-18

I have just tested it (gemini 3 pro thinking) , it is the first model that pulled off a complex work related task. Will be testing on more work related tasks that all other models (including gemini 2.5) have been failing so far, and re-post. What a time to be alive!!!

95 5 replies
@IndigoGPT 2025-11-18

This is going to change the world! I can’t wait for NanoBanana 2

57 6 replies
@TheHistoryCode125 2025-11-18

The fact that it coded a working Minecraft clone in one shot with no external libraries is absolutely insane. The gap between this and previous models feels massive.

57 3 replies
@fstk9102 2025-11-18

what comes first, an LLM able to build GTA6, or actual GTA6 release?

31 2 replies
@thokozanimanqoba9797 2025-11-18

HERE WE GO!!! 🎉 Thank you for the news update finally. This will be like Bloomberg for AI news one day... team team team!!!

23 1 replies
@LeopardChick 2025-11-18

I’m anxiously awaiting for the new model to pop up on my pro account. 🎉🎉🎉🎉 Thanks Matt I look forward to your next video about Gemini 3.

17
@eric8357 2025-11-18

Nothing new about Nano Banana 2? I want to try it out.

6
@guru_singh_yt 2025-11-18

It's SWE score is same as Claude but REASONING lead makes it ONE SHOT so many things accurately so good, it's just amazing.

4
@wardehaj 2025-11-18

Great instructional video, thanks!

3
@cherrlyn381 2025-11-19

Thanks for keeping us up to date.

0

Unlock the Data Inside
Turn Videos into Knowledge

  • Get FREE 10/day: transcripts, summaries, chats
  • Chat with videos, export text & PDF
  • $1 free API credit for RAG, chatbots & research

Free forever plan • All features unlocked

App screenshot