Экстракт Знаний
The Week That Changes AI: China's FREE Model Beat GPT-5, OpenAI Hit 'Code Red' (+8 Insane Updates)
13:52

The Week That Changes AI: China's FREE Model Beat GPT-5, OpenAI Hit 'Code Red' (+8 Insane Updates)

Vaibhav Sisinty 07.12.2025 34 647 просмотров 1 000 лайков обн. 18.02.2026
Смотреть на YouTube
Поделиться Telegram VK Бот
Транскрипт Скачать .md
Анализ с AI
Описание видео
🔗 Join our WhatsApp Community Get the latest AI updates, tips, and insights straight to your inbox: https://dub.sh/ai-updates-vs This week in AI was wild. Sam Altman declared "Code Red" at OpenAI, pausing ads, health features, and shopping to focus on making ChatGPT better. Meanwhile, China's DeepSeek dropped two free models, with V3.2-Speciale outscoring GPT-5 on math benchmarks. Perplexity now offers Google's Nano Banana Pro, OpenAI's Sora 2 Pro, Claude Opus 4.5, and Kimi K2 all in one subscription. Google launched Workspace Studio for building AI agents in plain English. Amazon's Nova Act automates browser tasks with 90% reliability. ByteDance released Seedream 4.5 with better text rendering. Kling AI had five major announcements including Video 2.6 with native audio. Runway's Gen-4.5 is now the top AI video model. Krea AI lets you turn workflows into apps. And Tesla's Optimus robot is now running at potentially record-breaking speeds. Drop a comment and subscribe for next week's AI news! TOP 10 AI TOOLS: 1. Stickerbox (Screen-free AI for kids): https://stickerbox.com/ 2. Eleven Reader (AI text-to-speech): https://elevenlabs.io/reader 3. AppForceStudio (Build iPhone apps by chatting): https://appforcestudio.com/mobile-app 4. Cleva (Learn software without switching tabs): https://www.cleva.io/for-creators 5. X-Design 2.0 (Brand consistency tool): https://www.x-design.com/ 6. Gleam (AI design feedback): https://v1labs.com/gleam 7. NeuralAgent (AI that controls your computer): https://www.getneuralagent.com/ 8. Taskade Genesis (AI workflow builder): https://www.taskade.com/ 9. Protaigé (AI marketing agency): https://www.protaige.com/ 10. Superfill (Auto-fill applications): https://superfill.ai/ 0:00 - Intro 0:32 - Deepseek V3.2 Beats GPT-5 at Math 1:07 - Sam Altman Declares Code Red at OpenAI 1:25 - Perplexity Now Has Every Top AI Model 2:30 - Google Workspace Studio: Build AI Agents in Minutes 3:54 - AWS Nova Act: AI That Controls Your Browser 4:34 - ByteDance Seedream 4.5: Character Consistency Fixed 5:34 - Kling AI Omni Week: 5 Days, 5 Announcements 5:48 - Day 1: Kling O1 (Video Gen + Editing) 6:15 - Day 2: Image 01 6:34 - Day 3: Video 2.6 with Native Audio 7:05 - Day 4: Avatar 2.0 7:34 - Day 5: Element Library 8:19 - Runway Gen 4.5: New #1 AI Video Model 9:20 - Krea Node App Builder: Turn Workflows into Apps 10:25 - Tesla Optimus Running at 8.5 MPH 11:44 - Top 10 AI Tools This Week 13:37 - Outro -------- To Know More, Follow Vaibhav Sisinty On ⤵︎ Instagram @VaibhavSisinty https://www.instagram.com/vaibhavsisinty Twitter @VaibhavSisinty https://twitter.com/VaibhavSisinty Facebook @VaibhavSisinty https://www.facebook.com/vaibhavsisinty/ LinkedIn - Vaibhav Sisinty https://www.linkedin.com/in/vaibhavsisinty

Оглавление (18 сегментов)

  1. 0:00 Intro 74 сл.
  2. 0:32 Deepseek V3.2 Beats GPT-5 at Math 99 сл.
  3. 1:07 Sam Altman Declares Code Red at OpenAI 49 сл.
  4. 1:25 Perplexity Now Has Every Top AI Model 180 сл.
  5. 2:30 Google Workspace Studio: Build AI Agents in Minutes 263 сл.
  6. 3:54 AWS Nova Act: AI That Controls Your Browser 136 сл.
  7. 4:34 ByteDance Seedream 4.5: Character Consistency Fixed 155 сл.
  8. 5:34 Kling AI Omni Week: 5 Days, 5 Announcements 37 сл.
  9. 5:48 Day 1: Kling O1 (Video Gen + Editing) 87 сл.
  10. 6:15 Day 2: Image 01 53 сл.
  11. 6:34 Day 3: Video 2.6 with Native Audio 69 сл.
  12. 7:05 Day 4: Avatar 2.0 75 сл.
  13. 7:34 Day 5: Element Library 141 сл.
  14. 8:19 Runway Gen 4.5: New #1 AI Video Model 162 сл.
  15. 9:20 Krea Node App Builder: Turn Workflows into Apps 199 сл.
  16. 10:25 Tesla Optimus Running at 8.5 MPH 206 сл.
  17. 11:44 Top 10 AI Tools This Week 314 сл.
  18. 13:37 Outro 47 сл.
0:00

Intro

Sam Altman just declared code red at OpenAI. An open-source model from China just outscored GPT5 on math. Google just let you build AI agents in minutes with zero code. And Cling AI dropped five major announcements in 5 days. Video Gen will never be the same. The AI world went crazy this week. These 10 actually matter. Stick around for the tools section. There's one you absolutely can't miss. Let's get into it.
0:32

Deepseek V3.2 Beats GPT-5 at Math

China just embarrassed OpenAI. GPT5 was crushing every math test out there until this open-source model just scored higher. Deepseek just dropped two new models, V3. 2 for everyday work and V3. 2 Special for heavy reasoning. You can use Deepseek's V3. 2 Special to solve insane math problems and work through research papers. Basically, anything that needs serious thinking. Both the models are free to download, but if you want to use special through their API, December 15th is the deadline. So yeah, Deepseek is giving all the major models out there some serious competition. Sam Alman just
1:07

Sam Altman Declares Code Red at OpenAI

declared code red at OpenAI. He's stopping everything. Ads, health features, shopping, all the side stuff. The whole team is now focused on one thing, making chat GPT better. We did a full breakdown on this. What's happening, why Altman's worried, all of it. Links in the description. Plexity
1:25

Perplexity Now Has Every Top AI Model

just solved the biggest problem in AI. Instead of juggling multiple platforms, you now get the best of Google, OpenAI, Anthropic, and Moonshot in one place. Max subscribers now get Nano Banana Pro as the default image generator. That's Google's Gemini 3 Pro image model with perfect text rendering and up to 4K resolution. For video, Sora 2 Pro is now default. It comes with integrated audio. Both pro and max users now get Kim K2 thinking and Gemini 3 Pro. Kimmy K2 is privacy first with reasoning mode always active. Perfect for step-by-step problem solving and technical tasks. And if all this wasn't enough, the Max users also get Claude Opus 4. 5, Anthropic's flagship model built for demanding reasoning and agentic workflows. Plus 03 Pro, OpenAI's top tier reasoning model. So now you've got Google's best image and text model, OpenAI's best video and reasoning models, Anthropic's most powerful model, and Moonshot's trillion parameter beast, all on one subscription. This is how AI should work. But before we dive in, I share updates like these daily on my
2:30

Google Workspace Studio: Build AI Agents in Minutes

free WhatsApp community. Links in the description. Google just dropped something insane. It's called Workspace Studio, and it basically lets you build your own AI agent in minutes. So here's how it works. You just type what you want in plain English. something like when I get an email, check the tone. If it's negative, label it and draft a reply. And Gemini just builds it. The whole workflow with steps, triggers, everything. You don't need to code or learn all the technical stuff. Just describe the problem and Google AI figures it out. It works right inside Gmail, Drive, Sheets, Chat, all your workspace apps. And they've got a bunch of templates already made for you, like auto follow-ups, alerts for urgent emails, pinging your team every week. It's all just sitting there ready to use. Let's try this one out. Premeating briefs. Turn it on and you'll get a quick summary before every call. Who's there? What matters? What to do? But here's where it gets wild. You can also build from scratch. You just tell it what you want. Like, hey, when I get an email, check if it sounds angry. If it does, label it and write me a reply. Okay, check this out. We have an email from an angry customer. The agent picks up on the tone, writes an apology, and drops in a 30% discount code. You didn't touch anything. Google's calling it Gemini powered automation for everyday work. And honestly, this changes everything. What would you automate first? AWS just launched an AI that can
3:54

AWS Nova Act: AI That Controls Your Browser

control your browser and actually do stuff for you. It's called Nova Act. You tell it what you want in plain English and it takes over your browser and just does it like log into this site, pull the sales numbers and drop them in a spreadsheet and it figures out the clicks, the scrolling, everything. Now, you're probably thinking, "Yeah, I've seen these before. They break after 5 minutes. " This one's hitting 90% success rates at actual companies. Herz is using it and they're moving five times faster now. The difference is Amazon didn't just slap an AI onto a browser and call it a day. They trained the entire system together. The AI, the browser controls all of it, so it actually understands what it's doing. Bite Dance just dropped a stealth update
4:34

ByteDance Seedream 4.5: Character Consistency Fixed

that nobody is talking about. Cdream 4. 5 is live and it's solving the biggest headache in AI image generation. Cream 4. 5 excels at character consistency. It can now fuse up to 10 reference images and keep everything locked. Face, clothes, lighting, style. The text rendering is also insane. Small text, logos, posters, it actually reads clean now. And it just hit number seven on the textto image arena leaderboard. That's the public voting benchmark where real users pick their favorite outputs blind. What's wild, this isn't some major overhaul. Bite Dance calls it a refinement focused upgrade. Meaning they just made everything work better. Cinematic lighting, sharper details, better spatial reasoning. The kind of stuff that actually matters for e-commerce, film, game art, advertising. It's free to try right now on Bite Plus Model Arc. 200 images, no credit card. So, image gen is taken care of, but what about videogen? Clingai had
5:34

Kling AI Omni Week: 5 Days, 5 Announcements

one of the busiest weeks ever. They just wrapped up their omni week. 5 days, five announcements. Here's everything they dropped and whether any of it is actually worth your time. On day one, they dropped Cling01.
5:48

Day 1: Kling O1 (Video Gen + Editing)

It's a unified model where video generation and editing happen in the same place. It's crazy what all you can do with this thing. You can morph images, put yourself into any video, remove unwanted objects, swap faces, and change camera angles or art styles just by typing a prompt. We have done a detailed video on this where I tested it with nine increasingly difficult use cases. Video link is in the description. Check out to see how it did. Moving on to day two, they
6:15

Day 2: Image 01

introduced image 01. As the name suggests, this is an image gen model. It comes with one-click outfit swaps, turning floor plans into 3D renders, and much more. If you're on their Pro, Premiere, or Ultra plan, you get a year of unlimited access to it. Day three, and this was a major
6:34

Day 3: Video 2.6 with Native Audio

update. The only criticism that Clingo 1 was getting was their lack of audio. Cling shut the noise down with the announcement of video 2. 6. It generates audio with your video, dialogue, sound effects, ambient noise, all in one pass. Cling put out a short film called I have a secret to demo it. — I have a secret. too big secret that I never told you. Jesse
7:05

Day 4: Avatar 2.0

— day four was for Avatar 2. 0. This one's for creating AI spokespersons and presenters. You can generate full talking head videos up to 5 minutes long. The demo reel is kind of wild. The expressions are more natural than the usual AI avatar stuff. Less stiff, more like someone actually talking. It handles different languages, too. Korean, Chinese, English. They're pushing Avatar 2. 0 for ads, explainers, that kind of content. Day five
7:34

Day 5: Element Library

marked the last day of their Omni launch week, and they ended it with a bang. Cling launched their element library. You can now build your elements with images from multiple angles and have Clingo One remember your characters, items, and backgrounds for consistent results regardless of how the camera moves or how the scene develops. Alex Patrascu, the creative partner of Clingi, made an epic film using Cling, specifically the the element library and shared it on Twitter. The quality of this 6-minute film shows how good this Cling lineup is. So that's the full week, five updates, one platform. They've also got a creative contest running until December 14th. Top prize being a,000 bucks. Cling came out of nowhere and seized the video gen game. Now, who will take this crown next? I might know the answer to that.
8:19

Runway Gen 4.5: New #1 AI Video Model

Runway just released their new video gen model and it's insane. Runway Gen 4. 5 produces cinematic and highly realistic outputs while providing limitless creative freedom and precise control over every aspect of generation. It understands real world physics way better. Objects move with actual weight and momentum. Surfaces behave like they would in real life. Gen 4. 5 can follow complex prompts with camera choreography, scene timing, and atmosphere changes in one go. Texttovideo, imagetovideo, key frames, videotovideo. All coming to this model. 2 years ago, they launched Gen 1, the first public video generation model. Now, their latest model, Gen 4. 5, is the number one AI video model in the world. It scored 1,247 ALO on the artificial analysis leaderboard, beating every other AI video model in the world. It's rolling out now and will be available to everyone in the coming days. Runway Gen 4. 5 is now the new standard for AI videos. Korea just dropped something
9:20

Krea Node App Builder: Turn Workflows into Apps

huge. They are calling it node app builder and it lets you turn your workflows into simple apps that anyone can use. So, Crea Nodes lets you chain together 50 plus AI models, image generators, video tools, audio, 3D into one workflow. But now with Node App Builder, you can package any of those workflows into a clean app. There's a new button on the left sidebar of create nodes. Click it to activate the app builder. From there, you click on any property in your nodes to make it an input. Scene descriptions, file uploads, whatever you want users to control. And yeah, you can add multiple inputs, then you click the output of any node in your workflow to display it in the final app. Let's say you want to build a photo shoot app. You upload a product image, type a scene description like underground cinematic or snow mountain, and it generates insane product renders automatically. You can rename each input and output from the sidebar, and you choose who sees your app, public, private, or only visible to your teammates. Your workflow can now become apps for the whole world to use. In robotics, we
10:25

Tesla Optimus Running at 8.5 MPH

have some good news from Tesla. They just shared a video on Twitter where the Tesla Optimus is seen running. Now, some internet analysts have claimed this might be the fastest running humanoid robot at 8. 5 mph. This record is currently officially held by a Chinese robot called Robotera Star won at 8 mph. But here's the insane part. Just 2 and 1/2 years ago, this thing could barely walk. Now, it's potentially matching the world's fastest humanoid robot. Pilot production is already running at their Fremont factory. They're planning a bigger Gen 3 production line for 2026, and they're targeting a cost of just $20,000 per robot. What most people miss is that Optimus runs on the same tech as Tesla's cars, same batteries, same AI chips, same training data from billions of miles driven. A lot of people still think humanoid robots will never be practical. But when someone like Elon Musk is investing this heavily into it, maybe it's worth paying attention. And honestly, I think we're just getting started. Last month, Tesla shareholders approved Musk's trillion dollar pay package. And this is how they celebrated. The next few years in robotics are going to be insane. Now for our top 10
11:44

Top 10 AI Tools This Week

coolest AI tools that launched this week's segment where we look at the most fun and underrated tools. Let's go. Number one, Sticker Box is screenfree AI for children. You talk to it and it prints stickers. No phone, no screen, just a thermal printer that turns your voice into instant stickers. Number two, 11 reader reads anything out loud with supernatural AI voices, books, PDFs, articles, anything. Students and teachers get it free for a whole year. Number three, App For Studio lets you build real iPhone apps just by chatting on your phone. No laptop needed. Just describe your app and it builds it. Number four, Clever teaches you software without switching tabs. It watches your screen and shows you exactly where to click in Figma or Blender. No more pausing tutorials. Number five, XDesign 2. 0 keeps your branding consistent. It creates something called brand DNA. So your logo, packaging, and signs all look like they came from the same designer. Number six, Gleam roast your designs before you launch. It gives you feedback from 10 different AI experts like an accessibility checker or a harsh critic. Number seven, Neural Agent takes over your computer. It's not a chatbot. It actually clicks, types, and scrolls through your apps to finish tasks while you watch. Number eight, Taskade Genesis builds entire workflows from one sentence. You type what you need and it creates the whole system with AI agents already working. Number nine, Protage is like having a marketing agency in the cloud. You give it your brand rules and it creates the whole campaign with copy, strategy, and visuals. And number 10, Superfill. This one's wild. It remembers everything about you and fills out job applications, essays, anything. One click and it's done. Works offline and stays private on your device. Which one of these tools are you trying first? Let
13:37

Outro

me know in the comments. Open AAI is in panic mode. China keeps dropping free models that embarrass billiondoll labs. And the AI video war just hit a whole new level. Drop a comment with your thoughts and subscribe for next week's AI news. See you then.

Ещё от Vaibhav Sisinty

Ankur@warikoo Untold Podcast: From Employee to Founder to Creator 1:26:58

Ankur@warikoo Untold Podcast: From Employee to Founder to Creator

Vaibhav Sisinty | 07.10.2023 | 34 сегм. | 875 628
Google's SECRET 7 AI Tools Just DESTROYED ChatGPT (100% FREE Stack) 13:12

Google's SECRET 7 AI Tools Just DESTROYED ChatGPT (100% FREE Stack)

Vaibhav Sisinty | 12.11.2025 | 20 сегм. | 361 978
Google's FREE Tool Just DESTROYED 10 AI Subscriptions (NotebookLM Mastery) 16:26

Google's FREE Tool Just DESTROYED 10 AI Subscriptions (NotebookLM Mastery)

Vaibhav Sisinty | 23.12.2025 | 13 сегм. | 261 010
⁠Podcast with @Sahil_Bloom on Content Creation, Side Hustles & Fitness | GSTH 04 1:31:57

⁠Podcast with @Sahil_Bloom on Content Creation, Side Hustles & Fitness | GSTH 04

Vaibhav Sisinty | 20.04.2024 | 20 сегм. | 218 777
This Viral AI Bot Can do your Work For FREE | Here's How to Set It Up 12:32

This Viral AI Bot Can do your Work For FREE | Here's How to Set It Up

Vaibhav Sisinty | 29.01.2026 | 15 сегм. | 185 947
Master ChatGPT Agent Builder Before It's Too Late: Dev Day Breakdown + Full Tutorial 17:38

Master ChatGPT Agent Builder Before It's Too Late: Dev Day Breakdown + Full Tutorial

Vaibhav Sisinty | 08.10.2025 | 21 сегм. | 163 843
Ctrl+V

Экстракт Знаний в Telegram

Транскрипты, идеи, методички — всё самое полезное из лучших YouTube-каналов.

Подписаться