港股研究社讯，据彭博社消息，强脑科技目前正就以超13亿美元的估值筹集资金展开洽谈，后续或选择 在香港或中国内地进行首次公开募股（IPO）。 有知情人士透露，这家由哈佛校友韩璧丞于2015年创立的公司，正在洽谈获取约1亿美元的IPO前融资 事宜。 该知情人士还称，这家初创企业已着手准备上市相关文件，不过上市地点及其他细节尚未确定。目前， 有关融资和IPO的讨论仍存在诸多变数，具体细节可能会随市场状况的改变而调整。截至目前，强脑科 技的代表尚未对相关置评请求作出回应。 强脑科技由哈佛大学校友韩璧丞于2015年正式创办。该公司与埃隆·马斯克（Elon Musk）旗下的脑机接 口企业Neuralink在业务领域存在竞争关系。强脑科技和DeepSeek等公司一同被并称为杭州"六小龙"，其 主要业务聚焦于开发仿生肢体技术以及人脑控制计算机的相关技术。 ...

Core Viewpoint - The article discusses the recent decline in the usage and market share of DeepSeek, questioning the validity of the reported statistics and emphasizing the importance of considering third-party API usage in evaluating its performance [2][4][10]. Summary by Sections Market Share and Usage Statistics - Reports from Semianalysis indicate that DeepSeek's market share has dropped to below 5%, with a significant decline noted since January [4][10]. - The statistics cited by Semianalysis primarily focus on the official API usage, potentially overlooking significant third-party integrations and deployments [10][12]. Third-Party API Usage - DeepSeek's third-party API calls have reportedly increased nearly 20 times since the release of versions V3 and R1, indicating sustained interest from developers [11][12]. - The article argues that the decline in official API usage does not reflect the overall demand for DeepSeek, as many applications integrate it without being captured in the official statistics [10][12]. Comparative Performance - Data from OpenRouter shows that DeepSeek V3 has a tokens consumption of 378 billion, ranking it third behind Claude Sonnet 4 and ahead of Google’s Gemini [17][22]. - Despite a decline in market share, DeepSeek maintains over 50% of the domestic B-end demand, indicating its strong position in the market [33]. User Preference and Community Engagement - A survey by Artificial Analysis found that 53% of respondents still prefer DeepSeek, placing it fourth among AI product providers [39]. - DeepSeek-R1 continues to lead in popularity on platforms like Hugging Face, indicating strong community support despite market fluctuations [44]. Industry Context and Future Outlook - The rapid evolution of AI technology suggests that a decline in DeepSeek's market share may not indicate a loss of relevance but rather reflects the dynamic nature of the industry [49]. - The article highlights the importance of open-source contributions from DeepSeek in promoting AI equity, contrasting it with other companies that are moving away from open-source models [49][50].

Core Viewpoint - OpenAI has released its first open-source language model series, gpt-oss, which includes gpt-oss-120b and gpt-oss-20b, both of which are fully customizable and support structured output [2][3]. Model Specifications - gpt-oss-120b requires 80GB of memory to run, while gpt-oss-20b only needs 16GB [2]. - The models utilize a mixture of experts (MoE) architecture, activating 5.1 billion parameters per token for gpt-oss-120b and 3.6 billion for gpt-oss-20b, with total parameters of 117 billion and 21 billion respectively [9]. - Both models support a context length of up to 128k and are designed for efficient deployment on consumer-grade hardware [10]. Training and Performance - The training process for gpt-oss models combines reinforcement learning and techniques from OpenAI's advanced internal models, focusing on reasoning capabilities and efficiency [8]. - gpt-oss models have shown strong performance in reasoning tasks, with gpt-oss-120b performing comparably to OpenAI's proprietary models in core inference benchmarks [10]. Comparison with Competitors - Claude Opus 4.1 has demonstrated superior programming performance with a score of 74.5% in SWE-bench Verified programming evaluations, outperforming previous versions [5]. - Independent benchmark tests indicate that gpt-oss-120b is less intelligent than DeepSeek R1 and Qwen3 235B, although it has advantages in efficiency due to its smaller parameter size [13]. User Feedback and Limitations - Users have reported mixed experiences with gpt-oss models, noting that gpt-oss-120b is particularly unstable for coding tasks, while gpt-oss-20b performs better [6][17]. - The models exhibit a higher hallucination rate, with gpt-oss-120b and gpt-oss-20b generating hallucinations at rates of 49% and 53% respectively, significantly higher than OpenAI's previous models [16]. Open Source and Accessibility - gpt-oss models are released under the flexible Apache 2.0 license, making them accessible for various applications, including agent workflows and tool usage [11][10]. - The models are available for free download on Hugging Face, promoting wider adoption and experimentation within the developer community [2][3].

Core Insights - OpenAI has released two new open-source AI models, GPT-oss-120b and GPT-oss-20b, marking the first introduction of new open-source large language models since the release of GPT-2 in 2019 [1] - The release was initially planned for March but was delayed until August 6, following a global open-source movement sparked by DeepSeek earlier this year [1] - Both models are released under a permissive Apache 2.0 license, allowing businesses to use them commercially without prior payment or licensing [1] - OpenAI CEO Sam Altman described GPT-oss as a significant breakthrough, claiming it offers advanced open-weight inference capabilities comparable to o4-mini, and can be run locally on personal computers or smaller devices [1]

"XX发布最强开源大模型，多项基准测试全面超越XX等闭源模型！" "万亿参数开源模型XX强势登顶全球开源模型榜首！" "国产之光！XX模型在中文评测榜单拿下第一！" 随着AI时代的到来，各位的朋友圈、微博等社交平台是不是也常常被诸如此类的新闻刷屏了？ 今天这个模型拿到了冠军，明天那个模型变成了王者。评论区里有的人热血沸腾，有的人一头雾水。 一个又一个的现实问题摆在眼前： 这些模型所谓的"登顶"比的是什么？谁给它们评分，而评分的依据又是什么？为什么每个平台的榜单座次都不一样， 到底谁更权威？ 如果各位也产生了类似的困惑，说明各位已经开始从"看热闹"转向"看门道"。 本文之中，我们便来拆解一下不同类型"AI竞技场"——也就是大语言模型排行榜——的"游戏规则"。 01 类型一：客观基准测试（Benchmark），给AI准备的"高考" 人类社会中，高考分数是决定学生大学档次的最主要评判标准。 同样地，在AI领域，也有很多高度标准化的测试题，用来尽可能客观地衡量AI模型在特定能力上的表现。 因此，在这个大模型产品频繁推陈出新的时代，各家厂商推出新模型后，第一件事就是拿到"高考"考场上跑个分，是 骡子是马，拉出来遛遛。 ...

Core Viewpoint - The article highlights the proactive approach of Wuhan's citizen service hotline "12345" in addressing urban issues before residents even voice their concerns, showcasing a shift from reactive to proactive governance through data analysis and technology [1][3][4]. Group 1: Proactive Governance - The "12345" hotline has implemented a system that analyzes citizen complaints to predict and resolve issues before they are reported, leading to over 20,000 preemptive actions in the first seven months of the year [1][3]. - The service has successfully reduced the number of complaints by 28% during the period when proactive alerts were issued, indicating the effectiveness of the "未诉先办" (preemptive action) strategy [3]. Group 2: Data-Driven Decision Making - The "灵醒" platform utilizes extensive historical and real-time data to monitor urban issues, allowing for a shift from reactive "firefighting" to proactive "fire prevention" in city management [4][5]. - The platform's new modules, such as "民生十二时辰" and "民情月历," help track and analyze the frequency and trends of citizen complaints, enhancing the ability to anticipate and address issues [4]. Group 3: Technological Integration - The integration of AI technology, specifically the "灵醒AI助手," has improved the efficiency of the hotline, allowing for rapid and accurate responses to citizen inquiries, with a 92.7% accuracy rate in dispatching information [6]. - The AI assistant has transformed the hotline's operations, enabling quicker responses and reducing the time taken to fill out service requests, thereby enhancing citizen satisfaction [6].

Core Insights - OpenAI has released two open-weight language models, gpt-oss-120b and gpt-oss-20b, marking its first open-weight model launch since 2019 and responding to competition from Meta, Mistral AI, and DeepSeek [1][2][12] Model Specifications - gpt-oss-120b and gpt-oss-20b are designed for low-cost options, with gpt-oss-20b able to run on a laptop with 16GB RAM and gpt-oss-120b requiring approximately 80GB RAM [2][5] - gpt-oss-120b has a total of 117 billion parameters, activating 5.1 billion parameters per token, while gpt-oss-20b has 21 billion parameters, activating 3.6 billion parameters per token [5][6] Performance Evaluation - gpt-oss-120b performs comparably to OpenAI's o4-mini in core inference benchmarks, while gpt-oss-20b matches or exceeds the performance of o3-mini [7][8] - Both models utilize advanced pre-training and post-training techniques, focusing on efficiency and practical deployment across environments [5][11] Security Measures - OpenAI has implemented extensive security measures to prevent malicious use of the models, filtering harmful data during pre-training and conducting specialized fine-tuning for security assessments [11] - The company collaborates with independent expert groups to evaluate potential security risks associated with the models [11] Market Impact - The release of these models is seen as a strategic shift for OpenAI, which had previously focused on proprietary API services, now responding to competitive pressures in the open-weight model space [12][15] - OpenAI has partnered with major cloud service providers like Amazon to offer these models, enhancing accessibility for developers and researchers [3][11]

Open AAI just announcing two new openw weight AI models. Those are models where some of the parameters around how they're trained are public and accessible. McKenzie Sagalos is here.She's got more on how this could change the AI race in today's tech check. And Mac, we're talking about not completely open source, but a little bit. And that's a really important distinction here, Becky.So, OpenAI is shifting strategy today, making its tech more accessible than it's been in six years. Because until now you coul ...

OpenAI is releasing a pair of open and freely available AI models months after China's DeepSeek found success with a similar approach https://t.co/6AzFhFInXK ...

Core Insights - OpenAI has released two open-weight language models, gpt-oss-120b and gpt-oss-20b, marking the first release since GPT-2 in 2019, aimed at providing lower-cost options for developers and researchers [1] - Open-weight models have publicly available parameters, offering transparency and control, but differ from open-source models which provide full source code [2] - OpenAI collaborated with major tech companies like Nvidia and AMD to ensure compatibility of the models across various chips [3] Industry Context - The release of open-weight models by OpenAI contributes to a growing ecosystem, with other companies like Meta and Mistral AI also launching similar models [2][3] - Nvidia's CEO highlighted OpenAI's role in advancing innovation in open-source software, indicating a significant impact on the AI landscape [4]