Ai2 introduces Tulu 3 405B, an open-source AI model with 405 billion parameters, surpassing DeepSeek’s V3 and OpenAI’s GPT-4o in key benchmarks.
Mistral AI has launched Mistral Small 3, a 24-billion-parameter open-source AI model that rivals larger models like Meta’s Llama 3.3 70B in performance while being faster and cheaper. Designed for on-premises deployment, it targets industries like finance and healthcare, offering advanced AI capabilities without the high costs of Big Tech solutions.
Alibaba has introduced the Qwen 2.5-Max AI model, claiming it surpasses DeepSeek’s V3 and other leading AI systems, highlighting rapid advancements in the field.
Discover 6 key ways China’s low-cost AI model DeepSeek is disrupting global markets and reshaping the economics of AI development.
Reliance Industries plans to build a 3-gigawatt AI data center in Jamnagar, India, costing $20–30 billion. The facility will use renewable energy and Nvidia chips, positioning India as a global AI contender.
Meta plans to invest $65 billion in AI infrastructure by 2025, targeting 1.3 million GPUs and a Louisiana data center to power its Llama 4 model, as global competition intensifies with Trump’s $500B Project Stargate.
Perplexity assistant for Android offers advanced task automation and app integration for free, disrupting the market dominated by costly alternatives.
OpenAI’s Operator uses GPT-4o to automate pizza orders, NBA ticket bookings, and image-based shopping. Available to U.S. Pro users, it faces EU delays and privacy concerns over 90-day data retention.
ByteDance’s Doubao-1.5-pro AI model surpasses OpenAI’s o1 in complex reasoning benchmarks, while UI-TARS automates computer tasks. Both models offer radical cost savings, reshaping global AI economics.
