Strengthening User-Customized Features and Service Innovation Based on High-Performance Models
Ecosystem Expansion Through Agent Efficiency Optimization and Large-Scale Infrastructure Investment

Key Headlines and Service Launches

OpenAI significantly enhanced personalization features — launched "Your Year with ChatGPT" summarizing users'' annual usage patterns in some regions; added tone personalization settings for conversation enthusiasm, warmth, and emoji usage. New high-performance model competition: Z.AI released GLM-4.7 (open-source model specialized for coding and complex reasoning); MiniMax released M2.1 on Kilo platform (surpassing existing models in speed and efficiency); Xiaomi released MiMo-V2-Flash globally (emphasizing reasoning and coding capabilities as general-purpose assistant). Design and document efficiency tools: Manus launched Manus Design View (precise design editing via prompt and voice); Google NotebookLM introduced data table function (converting source information into structured tables) and Lecture format (generating ~30-minute AI lectures). Ecosystem expansion: Cursor acquired Graphite (developer portal company) accelerating integrated development platform construction; Anthropic released Bloom (open-source tool automatically evaluating AI model bias and sabotage, measuring alignment at human judgment level). Infrastructure: OpenAI signed approximately $100B data center agreements with Oracle and SoftBank; Meta committed $65B for 2026 AI infrastructure. The week''s theme: AI development competition is bifurcating into "model capability race" (each company releasing increasingly capable models) and "infrastructure race" (data center, chip, and power capacity becoming binding constraints) — companies without both competitive models AND infrastructure scale face strategic disadvantage.