• Horizon AI
  • Posts
  • New Alibaba Qwen 3 AI Model is Here 🔥

New Alibaba Qwen 3 AI Model is Here 🔥

How to use the new Qwen 3 locally

Welcome to another edition of Horizon AI,

Alibaba released the next generation of its open-source large language models, Qwen3, today — and experts are already calling it another breakthrough in China’s rapidly growing open-source AI space.

Let’s get started!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

  • Alibaba Launches New ‘Hybrid’ AI Reasoning Models

  • OpenAI Rolls Out New Shopping Features With ChatGPT Search Update

  • AI Tutorial: How to use the new Qwen 3 locally

  • AI Tools to check out

  • The Latest in AI and Tech 💡

  • AI Findings/Resources

AI News

ALIBABA

Alibaba Launches New ‘Hybrid’ AI Reasoning Models

Alibaba’s Qwen team has just launched a new series of open-source, multimodal large language models called Qwen3, which appear to rival the performance of proprietary models from companies like OpenAI and Google.

Details:

  • The Qwen3 family comprises eight models, ranging from 600 million to 235 billion parameters, offering both dense and mixture-of-experts (MoE) architectures, and supporting 119 languages.

  • The new models are “hybrid” models, meaning they can take time to “reason” through complex problems or answer simpler requests quickly, offering users the flexibility to control the thinking budget.

  • Their flagship model, Qwen3-235B-A22B, outperforms DeepSeek R1, o1, and even rivals Gemini 2.5-Pro on key third-party benchmarks like ArenaHard, while also surpassing o3-mini on the latest versions of AIME and BFCL, a test designed to evaluate a model’s reasoning ability.

  • That model isn’t publicly available yet; the largest public Qwen3 model, Qwen3-32B, reportedly rivals R1 and surpasses OpenAI’s o1 on several tests.

  • In addition to the two MoE models, Qwen3 also includes dense models at various scales: Qwen3-32B, Qwen3-14B, Qwen3-8B, Qwen3-4B, Qwen3-1.7B, and Qwen3-0.6B.

  • They are available for download under an “open” license on AI dev platform Hugging Face and GitHub.

This launch intensifies the competition in China's AI sector, with major players like Baidu and DeepSeek also escalating their AI efforts to close the gap with the top AI models—and doing so at an incredible speed.

TOGETHER WITH GENEVA

The Science City You Didn’t Know You Needed to Visit

Did you know that the World Wide Web was born in Geneva, Switzerland? Indeed, the first version of the Internet cropped up at CERN in 1989. Today the world-renowned center is home to the largest particle accelerator and to the CERN Science Gateway – a must-see hub for science enthusiasts that features hands-on exhibits, immersive virtual reality experiences, and live demonstrations.

OPENAI

OpenAI Rolls Out New Shopping Features With ChatGPT Search Update

OpenAI is making ChatGPT a better shopping assistant; it can now show handy product cards with images, prices, ratings, and detailed descriptions for certain shopping-related queries.

Details:

  • All recommendations are organic, with OpenAI partnering with retailers to ensure up-to-date pricing.

  • Clicking on a product card opens a sidebar with more details, like where to buy, user reviews, and an option to ask further questions about the product.

  • OpenAI says it also plans to soon integrate its memory feature with shopping for Pro and Plus users, meaning ChatGPT will reference a user's previous chats to make highly personalized product recommendations.

  • The shopping updates are rolling out to all ChatGPT users, including free and paid accounts.

  • Other updates include improved citations, trending searches, and autocomplete. Additionally, OpenAI will now let you message or call 1-800-CHATGPT via WhatsApp.

The update comes as the search feature has become one of ChatGPT's fastest-growing and most-used capabilities, according to the company, which just reported that the function has crossed a major milestone, processing over one billion web searches in a single week.

AI Tutorial

How to use the new Qwen 3 locally

  1. Download LM Studio 

  2. Search for "Qwen 3".

  3. Select the size you want: 4B, 8B, 14B or 32B (Keep in mind your computer's capabilities when choosing the model size.)

  4. Click Download.

Once downloaded:

  1. Click on "New Chat", then choose "Select a model to load".

  2. Load the model you just downloaded and start chatting.

AI Tools to check out

🧠 Pieces: AI-powered Long-Term Memory for developers, capturing everything you work on across desktop apps, IDEs, browsers, and terminals.

🤖 Lindy: Build AI agents in minutes to automate workflows, save time, and grow your business.

📰 MGX: A multi-agent AI platform based on real software SOPs. Chat with the AI team leader, architect, engineer, and data analyst 24/7 to create websites, blogs, games and more.

🌐 Univerbal: Language learning with a conversational AI Tutor.

🌊 Video Ocean: Create stunning videos from text and images in minutes.

AI Findings/Resources

🤔 Does RAG make LLMs less safe? Bloomberg research reveals hidden dangers

💼 Major survey finds most people use AI regularly at work – but almost half admit to doing so inappropriately

🌍 Trust in artificial intelligence: global insights 2025

💡 DeepSeek’s success shows why motivation is key to AI innovation

👀 How AI is being used to create sophisticated scams that leave even experts second-guessing

The latest in AI and Tech

The company will "gradually stop using contractors to do work that AI can handle," according to co-founder and CEO Luis von Ahn, who also announced that the company will be "AI-first."

Peter Fitzgerald, Google's vice president of platforms and device partnerships, admitted during court testimony that Google started paying Samsung in January this year.

According to a Pew Research Center survey, about half of U.S. adults expect AI to worsen the quality of news over the next 20 years, while only 10 percent anticipate a positive effect.

After The Wall Street Journal reported that Meta AI was getting explicit with underage users, TechCrunch reported that ChatGPT seems to have the same issue: the site used a string of accounts aged 13 to 17 to ask the chatbot to "talk dirty to me," and the AI only said no once.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

We'd love to hear your thoughts on today's email!

Your feedback helps us improve our content

Login or Subscribe to participate in polls.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻