半导体：豆包新增视觉模型，SOC+CIS+存储有望重点受益

Investment Rating - The industry rating is "Outperform Market" (maintained rating) [1] Core Viewpoints - The release of the new visual understanding model by Doubao is expected to significantly enhance image and text recognition capabilities, simplifying application development processes and unlocking more value scenarios for large models [1] - The average input price for the Doubao visual understanding model is set at 0.003 yuan per thousand tokens, which is an 85% reduction compared to the industry average, allowing for the processing of 284 images at 720P for just one yuan [1] - The daily usage of Doubao's large model has exceeded 4 trillion tokens in December, representing a growth of over 33 times since its release in May [1] - The global smart toy market is projected to grow to nearly 70 billion USD by 2026, driven by the rise of generative large models and AI applications [1] - The SoC market is expected to grow from 138.46 billion USD in 2024 to 205.97 billion USD in 2029, with a CAGR of 8.3% [1] Summary by Sections Industry Rating - The industry maintains an "Outperform Market" rating, indicating a positive outlook compared to the broader market [1] Key Developments - Doubao's new visual understanding model enhances capabilities in various sectors including finance, healthcare, and logistics, with significant potential for application expansion [1] - The introduction of higher initial traffic for the visual understanding model aims to facilitate innovation and application development [1] Market Trends - The AI and machine learning demand is driving the SoC market, with a notable increase in the value of mainboard chips due to technological advancements [1] - The report suggests focusing on companies involved in AI glasses, home AI SoCs, visual SoCs, and AI toys, highlighting specific companies for potential investment opportunities [1]