In May 2025, the field of artificial intelligence has entered a deep integration phase across three major trends: large multimodal models, embodied intelligence, and computational infrastructure. Tech giants like Google, Microsoft, and NVIDIA are consolidating their leadership through new model releases and open ecosystem strategies. Meanwhile, companies like Anthropic and Baidu are making breakthroughs in complex task processing and cross-modal interaction. Tesla and Huawei are accelerating AI deployment in the physical world and enterprise-level management scenarios.
On May 13, Japanese startup Sakana AI released its new model CTM (Composable Thought Model), giving AI reasoning capabilities closer to that of the human brain. It uses a "step-by-step thinking" mechanism and demonstrates human-like interpretability and understanding in solving mazes and image recognition tasks—marking a significant milestone in embodied intelligence research.
On May 19, Microsoft launched the Windows AI Foundry platform, achieving an end-to-cloud AI development system. It integrates over 1,900 mainstream models and supports cross-device agent collaboration. GitHub Copilot was also upgraded to a “coding companion,” boosting development efficiency by around 40%. Thanks to deep collaboration with NVIDIA, Microsoft’s Azure platform—now integrated with H100 GPUs—has improved inference speeds by 50%. Microsoft also proposed the Model Context Protocol (MCP), natively integrated into Windows 11, laying the groundwork for future general-purpose agent ecosystems.
On May 20, Google unveiled the Gemini 2.5 Pro and Gemini 2.5 Flash models at its I/O 2025 conference. Gemini 2.5 Pro leads in math, programming, and logical reasoning, supporting multilingual voice and video comprehension, while Flash focuses on high efficiency and low cost. Google also introduced the AI Ultra subscription service, Veo 3 video generation tool, Flow AI video editing tool, and Imagen 4 image model—comprehensively revamping search, Gmail, and development tools, aiming to build a closed-loop AI “operating system.” The move is widely seen as Google's strong counter to OpenAI’s recent dominance.
On May 23, Anthropic hosted its “Code with Claude” developer conference, unveiling Claude Opus 4 and Claude Sonnet 4. Opus 4 specializes in complex reasoning and long-duration AI tasks, capable of running for up to 7 hours continuously—breaking the current limits of AI programming persistence. Sonnet 4 balances performance and cost. In the authoritative SWE-bench Verified benchmark, both achieved scores above 79%, surpassing GPT-4.1 and Gemini 2.5 Pro, solidifying Claude's leadership in software development.
May 28 became a “Super Tuesday” for AI.
Baidu launched the world’s first trillion-parameter multimodal model, “Wenxin·Lingmou,” with powerful voice, image, video, and text fusion capabilities. It boosts real-world comprehension and promises wide applications in finance, healthcare, transportation, and security.
On the same day, Tesla’s Shanghai Gigafactory officially began mass production of its humanoid robot, Optimus 2.0. The first 1,000 units were delivered to the Yangshan Port smart terminal. Equipped with Tesla’s self-developed D1 AI chip (200 TOPS), the robot boasts 0.1mm precision, 50km daily inspection range, and a failure detection rate three times higher than human inspectors—setting new standards for industrial robotics.
Also on May 28, NVIDIA made waves at Computex 2025 by launching its next-gen training chip, the H1000, with 5 PFLOPS of FP16 performance and support for trillion-parameter model training. GPT-5’s training time has dropped from 7 days to just 48 hours. Microsoft Azure and Alibaba Cloud have already deployed the chip. NVIDIA also announced a new Shanghai office to deepen its presence in China.
May 2025 has been another landmark month for AI. While tech giants like Google, Microsoft, and NVIDIA continue to shape the future with new models and ecosystems, companies like Anthropic and Baidu are pushing the boundaries of task complexity and multimodal intelligence. Yet perhaps the biggest headlines came from Tesla in the U.S. and Huawei in China.
What breakthroughs will June bring? Stay tuned at: https://iaiseek.com/