Xiaomi jumps into the AI race with its MiMo model

Xiaomi just unveiled MiMo, its first open-source artificial intelligence model, marking its official entry into the global AI competition. This 7-billion-parameter model, developed by the new "Xiaomi Big Model Core Team," stands out for its exceptional mathematical reasoning and code-generating capabilities.

• The model outperforms larger competitors like OpenAI's o1-mini and Alibaba's Qwen-32B-Preview on certain benchmarks, despite its modest size.

• MiMo has been pre-trained on approximately 25 trillion tokens and employs multi-token prediction to boost its performance.

• Xiaomi has reportedly invested heavily in its infrastructure, acquiring around 10,000 GPUs for training its AI models.

• The launch comes amidst increasing competition in the AI field in China, following recent advancements by DeepSeek and Alibaba.

🔗 Read the article

Facing ChatGPT and Gemini, Meta rolls out its Meta AI app

Meta has launched its own standalone generative AI application, stepping directly into the ring with OpenAI's ChatGPT and Google's Gemini. After integrating Meta AI into its existing apps (Facebook, Instagram, WhatsApp, Messenger), Mark Zuckerberg's company now offers a dedicated experience.

• The app is based on Llama 4, the latest language model developed by Meta.

• Meta AI leverages user profile data to customize the experience, but can also be used offline.

• Zuckerberg mentioned a potential premium version and the possibility of integrating ads and product recommendations.

• Meta AI already boasts nearly a billion monthly users across the company's app ecosystem.

  
  

Microsoft to integrate Grok into its Azure platform

Microsoft is set to integrate Grok, the AI model developed by Elon Musk's xAI, into its Azure cloud platform. Microsoft engineers have reportedly been instructed to prepare the necessary infrastructure to host this model.

• Grok will be available on Azure AI Foundry, Microsoft's AI development toolkit.

• This move comes in the context of tensions between Microsoft and OpenAI, its main AI partner.

• Microsoft is looking to diversify its AI model offerings, having already integrated DeepSeek's R1 model on its platform.

• The official announcement might take place at the Microsoft Build conference, scheduled for May 19, 2025.

🔗 Read the article

Apple ordered to pay half a billion dollars for patent infringement

The UK court has ordered Apple to pay $502 million to Optis for infringing 4G technology patents. This decision marks another legal setback for the California giant in its ongoing battle against what some call a "patent troll."

• The case dates back to 2017, when Optis accused Apple of using its essential patents without a proper license.
  The initial amount set in 2023 at $56.43 million was nearly multiplied by 10 following Optis' appeal.
  Apple criticizes a company that "doesn't develop any products" and solely relies on lawsuits.
  This ruling adds to recent penalties for patent violations, bringing the total to over $800 million paid to Optis.

🔗 Read the article

Alibaba launches Qwen 3, its next-gen AI models taking on global leaders

Alibaba has unveiled Qwen 3, its new family of AI models including eight variants, two of which are MoE (Mixture-of-Experts) models. The flagship model, Qwen3-235B-A22B, with 235 billion parameters, of which 22 billion are activated, directly competes with cutting-edge models from OpenAI and Google. Qwen 3 introduces the concept of "hybrid reasoning," allowing a switch between a "deep thinking" mode for complex tasks and a standard faster mode for general inquiries.

• Alibaba's first "hybrid reasoning" model

• Supports 119 different languages and dialects

• 300 million downloads for the Qwen series and over 100,000 derivative models

• Significant reduction in deployment costs compared to competing models

• Available open-source on Hugging Face and GitHub

🔗 Read the article

💡 What do you think of these developments? Opportunities for businesses or an AI-packed future invading our interactions?

At Cleo Academy, we train talents and companies to master these developments and make the most out of AI.
Want to know more about our programs?

Join us! 👉 www.cleo.academy

🚀 See you next Tuesday for another dose of tech and AI news!