June 2025 AI News Roundup: Breakthroughs, Surprises, and Global Developments
OpenAI CEO Sam Altman announced GPT-5 is expected to launch in summer 2025 and is reportedly materially better than GPT-4. Midjourney unveiled its first text-to-video system, Model V1, producing 16-second videos with advanced motion and style control in June 2025. China’s MiniMax released the M1 large model as an open-source Apache 2.0-licensed challenger claiming cutting-edge performance with far less computing power. Meta’s next-generation LLaMA 4 Behemoth was delayed to late 2025 due to performance issues. Google DeepMind introduced Gemini Robotics On-Device, a vision-language-action model that runs entirely on robots with on-board processing and can be fine-tuned with as few as