"当AI像电力一样流过，这个世界正在发生我们看不见的改变。 在科技圈，48小时足以完成一次行业认知的重塑。 2026年2月12日至14日，字节跳动在大模型领域发动了一场极具爆发力的"全线空袭"。2月12日，豆包视频生成模型Seedance 2.0率先登场，主打物理规律 的精准遵循；13日，图像创作模型Seedream 5.0 Lite接棒，首次引入了打破知识时效性局限的"实时检索"能力。 到了14日情人节当天，字节掏出了真正的压轴大戏——豆包大模型2.0（Doubao-Seed-2.0）。这也许就是AI时代独特的"浪漫"。 这种高频且密集的节奏，不仅是模型能力的代际跨越，更是字节跳动在2026年度关键词"勇攀高峰"下发起的一场饱和攻击。 外界最为好奇的是：为什么在激烈的模型竞赛中，字节能跑出"日均使用量突破63万亿Tokens"这一恐怖的数据？答案并不只在模型本身，而藏在"AI大模 型"与"火山引擎云底座"的深度耦合之中。 字节正在用豆包的"尖刀"技术，为火山引擎开辟出一片从未有过的AI云疆域。 大模型正处于一个微妙的"断档期"：PPT里的前景无限，对比着现实中的落地维艰。从2024年至今，AI行业正经历着一 ...

Group 1 - The State Administration for Market Regulation opposes manufacturers engaging in "involution" and selling cars at a loss, aiming to promote healthy competition in the automotive market [3] - The newly released guidelines clarify the boundaries of pricing behavior in the automotive industry, encouraging compliance and fair competition among manufacturers and sellers [3] - The guidelines specify that automotive manufacturers must not engage in pricing behaviors that aim to eliminate competitors or monopolize the market, which poses significant legal risks [3] Group 2 - OpenAI announced the release of a new real-time programming model, GPT-5.3-Codex-Spark, which is a more compact version of GPT-5.3-Codex, optimized for low-latency hardware [5][7] - The Codex-Spark model can generate over 1000 tokens per second and is currently available as a research preview for ChatGPT Pro users [5][7] - The model operates on Cerebras' Wafer Scale Engine 3, which features 40 trillion transistors and 900,000 AI cores, providing extremely low latency [7] Group 3 - Multiple automotive companies, including Geely, Chery, and BYD, have disclosed their plans for solid-state battery technology, with Geely aiming for small-scale production by 2027 and a long-term goal of mass production by 2030 [18][19][20] - Chery plans to start pilot production of solid-state batteries by 2026 and aims to validate the technology in real vehicles by 2027 [20] - BYD is focusing on sulfide solid-state batteries, expecting to achieve small-scale production by 2027 [23] Group 4 - Meizu's smartphone business is reportedly facing dissolution, with the Meizu 23 project allegedly halted [25][26] - The company has not officially responded to the rumors, and customer service has stated that no notifications regarding the business's status have been received [26] Group 5 - ByteDance is set to release the upgraded Doubao model on February 14, 2026, which will include significant enhancements to its foundational model capabilities [34][35] - The Doubao 2.0 model is expected to be a multimodal model with 1 trillion parameters, marking a substantial advancement in the company's AI capabilities [35] Group 6 - Xiaomi has launched its first-generation robot VLA model, which features 4.7 billion parameters and excels in visual language understanding and real-time execution capabilities [39][42] - The model has achieved state-of-the-art results in various benchmarks and is designed for high efficiency in real-world tasks [42] Group 7 - Lenovo reported a 72% year-on-year increase in AI-related revenue, which now accounts for 32% of the company's total revenue, driven by growth in AI PCs, smartphones, and servers [32][36] - The company has successfully navigated challenges in the global AI supply chain, maintaining its commitment to double-digit revenue growth and profitability [32]