## 🔍 Core Insights **Google DeepMind** rapidly launched its fourth Gemini 4 series model, **Gemini 4 Neon**, in about seven weeks, directly targeting Claude Opus 5.5 with a 1M token context and highly aggressive pricing [2][5][9]; meanwhile, data from China's AI office sector has been released, with **ByteDance's Doubao** Work DAU exceeding 8 million and **Tencent's WorkBuddy** DAU surpassing 10 million, as major tech companies intensify their competition over personal and enterprise workflows [1]. On the capital side, **Anthropic**'s trillion-dollar IPO narrative and $81.3 billion financing have sparked intense debate over its valuation logic [14][21]. ## 🚀 Key Updates - **Gemini 4 Neon surprise launch, crushing Opus 5.5 on cost-effectiveness** [2][5][9]: 1M context window, priced at $3 per million input tokens and $15 per million output tokens, ranked first on the AA Coding Agent Index. - **ByteDance and Tencent simultaneously disclose core AI office data** [1]: Doubao Work DAU exceeds 8 million, ARR over 2 billion; Tencent WorkBuddy DAU exceeds 10 million, daily token consumption over 100 trillion. - **Volcano Engine fully launches SeedRealtime 1.0 API** [6][11]: An end-to-end real-time audio-video call model that unifies voice, visuals, and timing, reducing audio-video conversation rhythm issues by half. - **Anthropic's trillion-dollar IPO sparks heated discussion; FT estimates it needs $602 billion in annual revenue** [4][14][21]: Based on its $1 trillion compute commitment and 40% operating margin, its revenue target already exceeds Alphabet and Apple. - **Arizona launches criminal investigation into OpenAI and other companies** [10]: Following the "rogue AI agent intrusion into banking system" incident, authorities are checking whether the companies involved have persistently violated state law. - **Nvidia tightens procurement guarantees for cloud providers; Jensen Huang's departure triggers AI stock sell-off** [20][24]: Guarantees for new cloud providers will only take effect after financing and acceptance; Japanese AI concept stocks were hit hardest, as the market shifts from "AI premium" to "AI discount." - **Tencent executive says US chip restrictions will slow China's model training** [13]: Liu Chiping stated that new chip licensing rules for overseas data centers may delay model releases, but inference can already run on domestic chips. - **Kunlun Wanwei's SkyProduction launches marketing studio** [16]: Integrates a creative case library, AI Agent, and visual canvas, covering commercial scenarios such as short-drama sales, TVC advertising, and UGC seeding. ## 🔗 Sources [1] ByteDance and Tencent simultaneously disclose core data — https://www.bestblogs.dev/article/403f028281?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [2] Just now, Gemini 4 Neon surprise launch, cost-effectiveness crushes Opus 5.5 overnight — https://www.bestblogs.dev/article/41612d755d?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item [4] FT estimates Anthropic needs about $602 billion in annual revenue to achieve a 20% return on its roughly $1 trillion compute commitment — https://aihot.news/items/qlevy5925xy06icp4w6mhrzfx [5] Google DeepMind releases Gemini 4 Neon, the fourth Gemini 4 model in about seven weeks — https://aihot.news/items/y31qy9