每经记者｜宋欣悦 每经编辑｜兰素英 5月29日凌晨，DeepSeek-R1-0528正式在Hugging Face平台开源。此前一日（5月28日），DeepSeek官方宣布DeepSeek-R1模型已完成小版本试升级，用户可 通过官方网页、App、小程序进行测试（打开深度思考），API接口和使用方式保持不变。 在此次更新中，模型代码能力的提升最为显著。知名代码测试平台LiveCodeBench显示，更新后的R1性能可以媲美OpenAI o3模型的高版本。 除代码能力外，R1新版本模型的文本理解与推理能力亦实现跨越式升级。其上下文长度拓展至128k，长文本提取的准确率也有显著提升。 《每日经济新闻》记者（以下简称"每经记者"）通过俄罗斯方块游戏开发、太阳系模拟与交互展示，以及设计英伟达最新财报网页这三项任务对R1新版本 进行了实测，发现DeepSeek-R1-0528的编程能力确实强大，对物理世界的理解以及文本的提取都相当准确。 知名AI评论人Haider在社交媒体上评价称，这是开源的一大胜利。 瑞士IT咨询公司CTOL Digital Solutions首席执行官Max Zhang则表示，在这一"小"进步之后 ...

目前该升级版的DeepSeek-R1-0528已经全量上线官方网页、APP、小程序等等，API也已经可以接入。 关于DeepSeek官方多么有诚意，我们已经在V3版本的升级上看到了——模型性能大幅提升只是开胃小菜，成本价格比 更是再度优化。这回的更新也是一样，新版本的DeepSeek-R1主要在编程能力上大幅提升。据一家LLM API接入网站 OpenRouter，这回的新版本R1的输入输出价格几乎与先前版本毫无变化！ | or DeepSeek: R1 0528 | | DeepSeek: R1 | (2) | | --- | --- | --- | --- | | Author | V deepseek @ | Author or deepseek @ | | | Context Length | 164K | Context Length | 164K | | May 28th update to the original DeepSeek R1 Performance | A | DeepSeek R1 is here: Performance on par with OpenAl o1. | 14 ...

Core Insights - Tencent's AI Workbench product, ima, has been updated to integrate the DeepSeek R1-0528 model, enhancing its capabilities for users [1] Group 1 - The update allows users to access the new DeepSeek R1 model without needing to update their existing ima versions, which includes desktop, app, and mini-program versions [1]

Group 1 - The A-share market experienced a significant rebound on May 29, with the Shanghai Composite Index closing at 3363.45 points, up 0.70% [1] - The A500 Index ETF (159351) saw a notable increase of 0.94%, with a closing premium of 0.13% [1] - The A500 Index ETF recorded a trading volume of 2.665 billion yuan, ranking second in the market for similar products, with a turnover rate of 18.26% [1] - Key stocks within the A500 Index ETF included Desay SV, which hit the daily limit, and Junshi Biosciences and Guangqi Technology, both rising over 7% [1] - The A500 Index ETF achieved a net subscription of 20.4 million shares, marking seven consecutive trading days of net inflows, with a year-to-date growth of 15.53%, the highest among similar products [1] Group 2 - The A500 Index ETF tracks the CSI A500 Index, comprising 500 stocks with large market capitalization and good liquidity, providing a balanced industry distribution and favoring large and mid-cap styles [2] - The ETF includes a high proportion of new productive forces, offering investors a tool to allocate to representative A-share companies [2] - Investors can access quality core asset opportunities through the A500 Index ETF linked funds (Class A 022453; Class C 022454) [2] Group 3 - CITIC Securities expressed an optimistic view on the current optimization of the Chinese capital market ecosystem, suggesting that the attractiveness of Chinese assets is on the rise [1] - There is an expectation for further stabilization of market confidence, with a more coordinated investment and financing environment and improved investor protection gradually forming [1] - Looking ahead, Chinese equity assets are anticipated to experience a significant market trend over the next year [1]

DeepSeek猝不及防地更新了，不是R2，而是R1v2。 官方的通告也很"DeepSeek"，甚至可以说有点佛系，在微信交流群发了一句： "DeepSeekR1模型已完成小版本试升级，欢迎前往官方网页、APP、小程序测试（打开深度思考），API接口和使用方式保持不变。" 看起来平平无奇，但真一顿实测下来，这个"小版本"恐怕还是过于谦虚了。 全球最大AI开源社区HuggingFace已经开源了这个新版本，名字也很实在：DeepSeek-R1-0528。不过截至目前，模型卡还没同步更新。 | deepseek-ai/DeepSeek-R1-0528 ♡ like 4 | | Follow or DeepSeek 69.9k | | | | | | | --- | --- | --- | --- | --- | --- | --- | --- | | & Safetensors deepseek_v3 custom_code | fp8 | | | | | | | | @ Model card Files and versions Community | | | | | | | | | main v DeepSe ...

规模方面，AI人工智能ETF最新规模达19.04亿元。 流动性方面，AI人工智能ETF盘中换手2.77%，成交5354.94万元。拉长时间看，截至5月28日，AI人工智能ETF近1年日均成交8769.66万元。 份额方面，AI人工智能ETF近1周份额增长900.00万份，实现显著增长，新增份额位居可比基金1/3。 资金流入方面，AI人工智能ETF最新资金流入流出持平。拉长时间看，近10个交易日内有6日资金净流入，合计"吸金"4201.15万元，日均净流入达420.12万 元。 数据显示，杠杆资金持续布局中。AI人工智能ETF最新融资买入额达209.72万元，最新融资余额达8464.96万元。 截至2025年5月29日 13:33，中证人工智能主题指数(930713)强势上涨1.90%，成分股德赛西威(002920)上涨10.00%，千方科技(002373)上涨5.76%，广电运通 (002152)上涨4.92%，四维图新(002405)，易华录(300212)等个股跟涨。AI人工智能ETF(512930)上涨2.05%，最新价报1.3元。拉长时间看，截至2025年5月28 日，AI人工智能ETF近1年累计上涨 ...

Core Viewpoint - The computer industry is experiencing a strong upward trend, driven by AI demand and policy support, with significant movements in stock prices and ETFs related to cloud computing and chips [3][4][5]. Group 1: Market Performance - The CSI Computer Theme Index rose by 1.93%, with notable gains in stocks such as Langxin Group (up 19.97%) and CloudWalk Technology (up 6.66%) [3]. - The Computer ETF (159998) increased by 2.25%, with a trading volume of 49.08 million yuan and a turnover rate of 1.73% [3]. - The CSI Hong Kong-Shenzhen Cloud Computing Industry Index saw a 1.49% rise, with Longbright Technology and Tianyuan Dike gaining 7.44% and 4.18%, respectively [3]. Group 2: Corporate Developments - On May 25, Zhongke Shuguang and Haiguang Information announced a merger plan to enhance business synergy and focus on AI full-stack solution development [3]. - The merger coincides with the revision of the "Major Asset Restructuring Management Measures for Listed Companies," indicating a new phase in optimizing industrial resource allocation [3]. Group 3: Investment Opportunities - The AI industry is expected to boost downstream demand in the computer sector, with a focus on AI computing power and domestic substitution trends [4]. - Investment strategies should consider the vertical integration capabilities of merged entities in cloud computing, which may enhance gross margins [4]. - The computer ETF has seen a significant increase in scale, growing by 23.02 million yuan over two weeks, and a notable inflow of 1.25 billion yuan over nine days [5].

Core Insights - DeepSeek has released a minor version upgrade of its R1 model, named DeepSeek-R1-0528, which has shown significant improvements in coding capabilities, nearly matching the performance of OpenAI's o3-high model [1][5] - Developers have noted enhancements in writing tasks, with outputs appearing more natural and better formatted compared to previous versions [7] - The model's performance in context recall has improved for contexts up to 32K, although there is a decline in performance for 60K contexts [7][8] Performance Metrics - In the Live CodeBench testing platform, DeepSeek-R1-0528 achieved a Pass@1 score of 73.1, ranking fourth among various models [3] - The top three models in the same test were 04-Mini (High) with 80.2, 03 (High) with 75.8, and 04-Mini (Medium) with 74.2 [3] Developer Feedback - Developers have expressed that the upgrade represents a significant victory for open-source initiatives [4] - Some developers have conducted personal tests comparing DeepSeek-R1 with Claude-4, finding R1 superior in certain aspects, such as the visual effects of a simulated collision [5] - There is anticipation for the next major version, R2, with hopes for improvements in context length and multimodal capabilities [8]

Core Viewpoint - The digital economy sector is experiencing significant growth, with key indices and ETFs reflecting strong performance, driven by advancements in AI and digital infrastructure [2][3][4]. Group 1: Digital Economy Index Performance - As of May 29, 2025, the CSI Digital Economy Theme Index (931582) rose by 1.11%, with notable gains from constituent stocks such as BGI Genomics (301269) up 13.85% and Winbond Electronics (600699) up 2.71% [1]. - The Digital Economy ETF (560800) increased by 1.09%, with a latest price of 0.74 yuan and a turnover rate of 0.61%, totaling a transaction volume of 458.85 million yuan [1]. - Over the past year, the average daily transaction volume for the Digital Economy ETF was 22.88 million yuan [1]. Group 2: AI Development and Market Trends - Guohai Securities expresses optimism regarding the accelerated development of AI, highlighting that companies proactively integrating AI are likely to see enhanced performance and valuation restructuring as market budgets shift from IT to human resources [2]. - The Digital Economy ETF closely tracks the CSI Digital Economy Theme Index, which includes companies with high digital infrastructure and application levels, reflecting the overall performance of the digital economy sector [2]. Group 3: Fund Size and Inflows - The Digital Economy ETF saw a significant increase in size, growing by 2.23 million yuan over the past month, ranking in the top half among comparable funds [3]. - The latest share count for the Digital Economy ETF reached 1.024 billion shares, marking a three-month high and also placing it in the top half of comparable funds [4]. - Recent net inflows into the Digital Economy ETF totaled 8.80 million yuan, with four out of the last five trading days showing positive net inflows, amounting to 14.75 million yuan in total [4].

Market Overview - A-shares experienced a broad rally, with the computer sector leading the gains, particularly the Xinchuang 50 ETF (560850) which surged over 2% and saw a net inflow of over 12 million yuan in the past five days [1][3] Company Developments - DeepSeek, a large model platform, released its latest version R1 on May 28 without detailed announcements, but it is expected to soon provide more information about its features. Testing by Live CodeBench indicates that its performance is comparable to OpenAI's latest models [3][4] - In the hardware sector, the Xinchuang 50 ETF (560850) component stocks, Haiguang Information and Zhongke Shuguang, announced a merger through a stock swap, with Haiguang being the second-largest holding in the ETF [3][4] Industry Insights - The Xinchuang 50 ETF (560850) tracks the Zhongzheng Xinchuang Index, which has a high weight of 48.1% in DeepSeek concept stocks, covering major players like Inspur Information and Kingsoft [7] - The index includes various sectors such as computer software (41%), cloud services (22%), computer equipment (19%), semiconductors (14%), and industrial internet (5%), indicating a strong linkage to the DeepSeek developments [7] AI Application Trends - AI companionship applications are gaining traction, with platforms like Character.AI and Talkie reaching over 10 million monthly active users. Key features include personalization, real-time interaction, and immersive experiences, primarily attracting younger demographics [5][6] - The integration of multi-modal capabilities in AI applications is expected to enhance user engagement and retention, with advancements in text, image, and audio generation technologies [5][6]