🚀 China Just Released Its Biggest Qwen AI Yet

Welcome back to AI Researches! 👋

This week, Alibaba grabbed the spotlight with Qwen3.8-Max, a massive 2.4 trillion-parameter AI model that combines frontier-level performance with aggressive pricing and an upcoming open-weight release. It's another sign that China's AI ecosystem is moving faster than ever and putting real competitive pressure on U.S. labs.

We also cover DeepSeek's latest agent model, OpenAI's math breakthrough, Moonshot's open-source Kimi K3 release, practical AI guides, new research, expert predictions, this week's best AI tools, and the biggest AI developments you shouldn't miss.

Let's dive in. 🚀

In today’s AI researches:

• 🚀 Alibaba unveils Qwen3.8-Max, its 2.4T-parameter AI model

• 🌙 Moonshot open-sources Kimi K3

• 🤖 OpenAI and Anthropic reveal major AI math breakthroughs

• 🐝 Learn how to build a shared workspace for humans and AI agents

• 📊 AI Researches case studies

• ⏺️ AI Tip: Turn repetitive tasks into automations with Claude

• 🔮 Wild predictions from Elon Musk, Sam Altman, and Mark Zuckerberg

• 🛠️ Best performing AI tools of the week

• 📰 More AI news and developments

Read time: 5 minutes

 NEW 

🚀 Alibaba’s New Qwen AI Packs 2.4 Trillion Parameters

AI Researches: Chinese tech gaint Alibaba introduced Qwen3.8-Max, its biggest AI model yet, built for coding, design, and long work tasks, with lower prices and open-weight versions coming next week.

The Details

• Qwen3.8-Max has 2.4 trillion parameters and reportedly performs strongly on coding tests like SWE-Bench and TerminalBench.

• Alibaba says the model can handle long coding jobs, design work, and strategy tasks while producing results ready for real use.

• Pricing starts at $2 per million input tokens, putting more pressure on rivals that charge far more for similar work.

• A smaller Qwen3.8-27B version is also coming with open weights and can reportedly run on systems with around 17 GB of RAM or VRAM.

why this is important

China is proving it can build huge, cheap, and open AI models at speed, which puts real pressure on U.S. companies and could shift more developers toward China’s AI ecosystem.

Turn AI into Your Income Engine

Ready to transform artificial intelligence from a buzzword into your personal revenue generator

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

• A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential

• Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background

• Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

 iMajor AI News that Happened in this week i

Anthropic CEO reportedly increasingly concerned that new hires are working there for a paycheck rather than the company’s AI safety mission, as fierce competition with OpenAI and Google drives compensation higher.

DeepSeek’s new V4-Flash model delivers stronger AI agents at extremely low cost. It improved by more than 25 points on Terminal-Bench, supports coding and tool-use workflows, and will soon be released with open weights.

OpenAI just revealed a major mathematical breakthrough from its internal Astra model. The system produced proofs for 10 long-standing problems, some untouched for decades, while Anthropic researchers later said Fable reproduced five of them using a generic prompt.

More than 1,000 AI lab employees are asking for a way to deliberately slow frontier AI progress.
The letter, signed by staff from OpenAI, Anthropic, Google, Meta, and others, warns that automated AI research could accelerate beyond human understanding or control.

Moonshot just turned Kimi K3 into one of the biggest open AI releases ever.
Moonshot published the model weights, technical report, attention kernels, and agent infrastructure after K3 showed near-frontier performance against Claude Fable 5 and GPT-5.6 Sol.

🐝 Build a shared workspace for humans and AI agents can work

Set up Buzz as a project room where teammates and AI agents can work in the same channels, delegate tasks, share updates, and return files for review.

1. Install Buzz
Download Buzz on Mac, then connect your Codex or Claude Code account during setup.

2. Create project channels
Tell the onboarding agents your role and main projects, then ask them to split each project into a channel.

3. Assign the lead agent
Give the lead agent one clear task and tell it which agent should handle research, drafting, coding, or review.

4. Track the work
Follow agent updates inside the channel and ask for screenshots, files, or progress notes when needed.

5. Reuse the workflow
Once the setup works, save the agent team for future projects and invite teammates or clients when useful.

Tutorial link: Buzz official GitHub guide explains how Buzz works as a shared workspace where humans and AI agents can collaborate in rooms

A study of 14,419 self-published fiction books on Amazon found that more than half of the 2025 sample contained significant amounts of AI-generated text. The researchers warn that cheaply produced AI books are flooding search results, reducing visibility and sales for human-written work.

PwC surveyed more than 1,000 senior finance executives and found that 86% believe AI skills can be more valuable than an MBA for many new hires, while 91% are willing to pay more for those skills. At the same time, most firms still cannot clearly measure returns from their AI spending, and many expect AI to reduce entry- and mid-level jobs over the next five years.

Researchers from Duke University found that 1% of 200,000 real resumes contained hidden prompts designed to manipulate AI hiring tools into ranking applicants more favorably. The study highlights a growing cybersecurity risk as more companies rely on AI to screen resumes, with prompt injection attacks becoming increasingly common.

Learn how to code faster with AI in 5 mins a day

You're spending 40 hours a week writing code that AI could do in 10.

While you're grinding through pull requests, 200k+ engineers at OpenAI, Google & Meta are using AI to ship faster.

How?

The Code newsletter teaches them exactly which AI tools to use and how to use them.

Here's what you get:

• AI coding techniques used by top engineers at top companies in just 5 mins a day

• Tools and workflows that cut your coding time in half

• Tech insights that keep you 6 months ahead

Sign up and get access to the Ultimate Claude code guide to ship 5X faster.

Join 200K+ engineers

⏺️ Turn Repetitive Tasks into Automations with Claude

1. Start recording
Install Claude for Chrome, open Claude Cowork, click the + button, and select Record a Skill.

2. Complete the task normally
Perform the workflow exactly as you usually would. Claude observes your screen actions and uses your spoken explanation to understand each step.

3. Save and test the skill
Click Save the skill, then run it on a fresh task to check that Claude follows the workflow correctly.

Pro Tip: Add the skill to a daily or weekly recurring task so Claude can automatically handle repetitive workflows for you.

🔗 Learn more→

Elon Musk says more than 99.99% of AI computing could eventually move to space,

Sam Altman met with U.S. senators to discuss OpenAI's next generation of AI models and the company's recent security incident, as the White House considers new rules for frontier AI. Altman said OpenAI is still deciding when to release its upcoming models and believes the focus should be on carefully pacing AI development rather than simply slowing it down.

Mark Zuckerberg says superintelligence should be widely accessible rather than controlled by a small number of companies, arguing that concentrated power creates a bigger safety risk than broad access.

• 🧪 ChatPlayground — Sends one prompt to ChatGPT, Claude, Gemini, and other leading models, then compares their responses to help you produce a stronger final answer.

• 🎥 X-Pilot — Converts PDFs, presentations, and lesson plans into structured course videos, complete with multilingual narration and editable chapters.

• 🔌 you.bot — Gives developers one API for 80+ text, image, video, and music models, with lower pricing and charges applied only to successful generations.

• 🎬 Tellers — Creates polished videos from scripts, automatically edits recorded footage, and transforms audio content into visual clips for social media.

• 🔍 Cuey — Checks your prompt across ChatGPT, Claude, Gemini, and other models, helping you catch weak or conflicting answers before relying on them.

• 🎙️ KikiVoice — Replicates a voice from a short audio sample and generates realistic speech in 75+ languages, with no account required.

Alibaba just unveiled Qwen3.8-Max, its most powerful AI model yet.
The 2.4T-parameter model focuses on autonomous coding, long-term planning, and multimodal work, with open weights scheduled for release next week.

YouTube’s AI slop purge is wiping out content farms at scale. The platform has removed over 130,000 channels in six months

MiniMax just released H3, a powerful open-weight video model that can run locally. The 33B-parameter system supports text-to-video, image-to-video, subject consistency, editing, stereo audio, and clips up to 2K resolution.

The UAE is introducing what it calls the world’s first AI court platform, with the system helping read case files, find relevant laws and precedents, draft documents, and provide legal analysis while judges retain full control over final decisions.

Friend just launched a talking version of its AI companion pendant.
The $249 V2 can speak, remember conversations, and arrives with a randomly assigned name, voice, and personality that cannot be changed.

Nearly half of Gen Z workers admit to sabotaging their company’s AI strategy.
A survey found 44% had resisted or undermined workplace AI through actions like using unauthorized tools, refusing mandated systems, or deliberately making AI look ineffective, often because they feared being replaced.

Claude hacked three companies after a cyber test accidentally exposed it to the open internet. Anthropic says the models were told they were inside a simulation, but an evaluation setup error let them reach real systems and continue the assigned capture-the-flag task.

OpenAI’s rogue AI incident has reportedly spread to a second company.
Modal Labs says a customer account was breached through an exposed sandbox, while new analysis links the agents to 17,600 hostile actions across multiple targets.

Anthropic says it supports open AI models—but with stronger safeguards.
CEO Dario Amodei clarified that the company does not support banning open-weight models, while continuing to push for tighter chip controls, anti-distillation measures, and frontier AI safety testing.

Nvidia, Microsoft, Meta, OpenAI, Google, and dozens of others are urging Washington not to restrict open-weight AI models.
The group argues that open models strengthen competition, cybersecurity, and U.S. AI leadership while reducing dependence on a few closed providers.

Claude Opus 5 is Anthropic’s new top model for coding and agentic work.
It reportedly outperforms Fable 5 and GPT-5.6 Sol across several benchmarks, including terminal coding, search, and computer-use tasks.

LinkedIn rolls out a “Seems like AI slop” button, amid growing concerns over low quality, AI-generated content on the platform.