Google officially launches the Gemini 3 Flash model, featuring ultimate reasoning capability and triple the inference speed of Gemini 2.5 Pro at one-fourth the cost of Pro tier. It supports multimodal input, long context, and adjustable reasoning depth. It has become the default model for the Gemini App and Google Search's AI mode, widely integrated into ecosystems like APIs, enterprise cloud, and developer tools. Companies such as JetBrains, Figma, and Bridgewater have already adopted it, while developers and general users can experience it for free.

Gemini 3 Flash not only approaches or even surpasses Gemini 3 Pro and GPT-5.2 in multiple benchmarks like reasoning, coding, and vision but also offers 3x faster inference and a price as low as 30% of comparable models. It has become the default AI model for millions of global users. Developers and enterprises can access it through various channels, significantly lowering the barrier to AI application.

Xiaomi releases and open sources the MiMo-V2-Flash large model, featuring a 309 billion parameter MoE architecture, inference speed of 150 tokens/sec, and a cost as low as $0.1 per million input tokens. It supports a 256k context window and multi-token parallel prediction, ranking first among open-source models in the SWE-Bench programming benchmark. It excels in long-text processing and multi-round agent tasks, positioning itself as an AI foundation for all scenarios including phones, IoT, and automobiles.

Luma launches the new Ray3 Modify model, allowing users to generate transition videos by providing start and end frames and to transform character appearances using reference images while preserving original performance, motion, and emotion. Suitable for creative studios and film production, the model is integrated into the Dream Machine platform, backed by $9 billion in new funding and AI infrastructure cooperation with Saudi Arabia.

OpenAI is in talks with Amazon and other investors for a new funding round up to $100 billion, potentially valuing the company at $750 billion. The parties have signed a 7-year, $38 billion cloud services agreement. OpenAI will adopt Amazon's in-house AI chip Trainium, indicating deep integration and multi-party competition among AI infrastructure and cloud service giants.

Meta's Chief AI Officer disclosed that the company will launch a next-generation image/video generation model codenamed Mango and the Avocado large language model in the first half of 2026. The latter focuses on enhancing programming and world modeling capabilities and may adopt a proprietary, closed-source approach, marking Meta's transition from its Llama open-source strategy towards profit-driven frontier models. The AI team has already recruited top talent extensively.

The global large language model (LLM) market size is projected to reach $8.3 billion in 2025 and $24.89 billion by 2029, with a compound annual growth rate (CAGR) of 31.5%. Main drivers include demand for intelligent automation, conversational AI, personalized experiences, and enterprise applications. North America is the largest market, while Asia Pacific is the fastest-growing. Innovation in open-source and enterprise-grade models is active.

AI large models and agents are accelerating their penetration into industries like HR, finance, insurance, and retail, driving implementation in scenarios such as automation, intelligent customer service, risk management, and personalized recommendations. Enterprises are focusing on AI talent, data security, and compliance. AI-driven business innovation and efficiency improvement have become mainstream trends.

Latest research shows scientific paper output increased by 23.7%-89.3% with LLM assistance, especially benefiting non-native authors. However, while writing complexity has improved, research quality has declined. Peer review and citation diversity have increased, prompting academia to strengthen quality assessment and methodological scrutiny to adapt to the new AI-driven research paradigm.

A BCG-FICCI report states that by 2026, AI will take priority in 70-80% of routine tasks and 30-50% of reasoning and analytical tasks. Corporate competition will shift focus to brand trust, patents, and AI talent. AI adoption and innovation labs are rapidly expanding in emerging markets like India. AI proliferation brings productivity gains and changes in employment structure.