China's FREE AI Just DESTROYED GPT-5 + OpenAI Declares 'Code Red' (+8 Shocking Updates)

Deepseek's Math Prowess and OpenAI's "Code Red" [00:32]

• A new open-source model from China has reportedly scored higher than GPT-5 on math tests, challenging OpenAI's dominance.
• Deepseek has released two new models: V3.2 for general tasks and V3.2 Special for demanding reasoning, both free to download.
• Sam Altman has declared a "code red" at OpenAI, pausing work on ads, health, and shopping features to solely focus on improving ChatGPT.

Basically, anything that needs serious thinking.

Plexity: A Unified AI Platform [01:25]

• Plexity aims to solve the problem of juggling multiple AI platforms by integrating the best models from Google, OpenAI, Anthropic, and Moonshot.
• Max subscribers now get Google's Gemini 3 Pro as their default image generator with high-resolution text rendering.
• Sora 2 Pro with integrated audio is the default video generator for Pro and Max users.
• Users gain access to Kim K2 for privacy-focused reasoning, Gemini 3 Pro, and for Max users, Claude Opus 4.5 and OpenAI's O3 Pro for complex reasoning and agentic workflows.

This is how AI should work.

Google Workspace Studio: Zero-Code AI Agents [02:35]

• Google's Workspace Studio allows users to build AI agents in minutes using plain English descriptions, with no coding required.
• These agents can automate workflows within Gmail, Drive, Sheets, and Chat, with pre-made templates available.
• An example demonstrated an agent that detects negative email tones, labels them, and drafts an apology with a discount code.

And honestly, this changes everything.

AWS Nova AI: Browser Automation [03:54]

• AWS has launched Nova AI, which can control a browser to perform tasks based on simple English commands, such as logging into sites or extracting data.
• Nova AI boasts a 90% success rate and has allowed companies like Herz to increase their operational speed fivefold.
• The system's effectiveness stems from the integrated training of the AI and browser controls, allowing for a deeper understanding of actions.

The difference is Amazon didn't just slap an AI onto a browser and call it a day.

ByteDance Cdream 4.5: Enhanced Image Consistency [04:34]

• ByteDance has released Cdream 4.5, focusing on character consistency in AI image generation by allowing up to 10 reference images to maintain elements like face, clothes, and style.
• Text rendering has significantly improved, making logos and text on posters appear clear.
• Cdream 4.5 achieved seventh place on the text-to-image arena leaderboard, a public voting benchmark.

Meaning they just made everything work better.

Cling AI's Omni Week: A Barrage of New Features [05:34]

• Cling AI held a five-day "Omni Week" with five major announcements, significantly impacting video generation.
• Cling01 unifies video generation and editing, enabling morphing, face-swapping, and style changes via prompts.
• Image 01 offers features like one-click outfit swaps and floor plan to 3D renders.
• Video 2.6 adds integrated audio generation, including dialogue and sound effects, demonstrated with a short film.
• Avatar 2.0 allows for the creation of realistic AI spokespersons and presenters for up to 5-minute talking head videos in multiple languages.
• The Element Library enables consistent character, item, and background reuse across scenes, as showcased in a 6-minute film by Alex Patrascu.

Cling came out of nowhere and seized the video gen game.

Runway Gen 4.5: The New Standard in AI Video [08:19]

• Runway has released Gen 4.5, which produces cinematic, realistic videos with high creative control and adherence to real-world physics.
• The model excels at complex prompts involving camera choreography, scene timing, and atmosphere changes.
• Runway Gen 4.5 is now considered the number one AI video model globally, scoring 1,247 on the artificial analysis leaderboard.

Runway Gen 4.5 is now the new standard for AI videos.

Kuaishou Node App Builder: Democratizing AI Workflows [09:20]

• Kuaishou's Node App Builder allows users to transform complex AI workflows, chaining together over 50 AI models, into simple, user-friendly apps.
• Users can define inputs and outputs for their workflows, enabling others to easily create specific outputs like product renders or photo shoots.
• These apps can be made public, private, or shared with teammates, making advanced AI accessible to a broader audience.

Your workflow can now become apps for the whole world to use.

Tesla Optimus: Rapid Progress in Robotics [10:25]

• Tesla's Optimus robot has been observed running at a potential speed of 8.5 mph, possibly surpassing the current record holder.
• In just 2.5 years, Optimus has evolved from basic walking to potentially matching the fastest humanoid robots.
• Pilot production is underway, with plans for a larger Gen 3 production line in 2026, targeting a cost of $20,000 per robot.
• Optimus leverages Tesla's existing automotive technology, including batteries, AI chips, and training data from billions of miles driven.

But when someone like Elon Musk is investing this heavily into it, maybe it's worth paying attention.

Top 10 Coolest AI Tools of the Week [11:41]

• Sticker Box offers screen-free AI for children, printing voice-generated stickers via a thermal printer.
• 11 Reader uses AI voices to read any text aloud, with a free year offered to students and teachers.
• App For Studio enables the creation of iPhone apps through chat interaction on a phone, no laptop required.
• Clever teaches software skills by watching the user's screen and providing visual guidance.
• XDesign 2.0 ensures branding consistency through a "brand DNA" feature.
• Gleam provides AI-driven design feedback from multiple expert perspectives.
• Neural Agent autonomously performs computer tasks by clicking, typing, and scrolling through applications.
• Taskade Genesis builds entire workflows from a single sentence, deploying AI agents.
• Protage acts as a cloud-based marketing agency, creating campaigns with copy, strategy, and visuals.
• Superfill remembers user data to complete job applications and essays with a single click, working offline and privately.

Which one of these tools are you trying first?