Group 1 - DeepSeek has announced the update of its official App, web version, and mini-program to DeepSeek-V3.2-Exp, resulting in a significant reduction in API costs by over 50% for developers [1] - The cost of AI inference computing power has been decreasing due to advancements in AI large models and improvements in the performance and cost-effectiveness of inference chips, with hardware costs dropping approximately 30% annually and energy efficiency improving by about 40% [1] - The continuous decline in costs for large models, represented by DeepSeek, supports the commercialization of AI applications and enhances the efficiency of distilled models [1] Group 2 - The rapid iteration of large models and enhanced inference capabilities are creating opportunities for customized Agent applications, allowing users to tailor agents based on personal data and needs [2] - Companies like Cambricon and Huawei Ascend have announced their compatibility with DeepSeek-V3.2-Exp and have open-sourced the vLLM-MLU inference engine [2] - Companies such as Fanwei Network, Kingsoft Office, and Dingjie Smart are involved in the development of Agent and AI applications [3] Group 3 - Huawei Ascend has achieved software and hardware adaptation with companies like Softcom and Changshan Beiming [4] - Jiuqi Software plans to upgrade its Nüwa GPT in early 2025, integrating deeply with mainstream large models and launching various intelligent applications [4] - Jiuqi Software's AI distillation technology is similar to that of DeepSeek, indicating a trend in the industry towards efficient model optimization [4]

协议要求加沙将由一个技术官僚、非政治性的巴勒斯坦委员会进行临时过渡治理，负责为加沙人民提供公共服务和市政的日常运行。该委员会将由符合条 件的巴勒斯坦人和国际专家组成，并由一个新的国际过渡机构"和平委员会"进行监督和监管，该委员会将由特朗普领导和主持，其他成员和国家元首将另 行公布。 每经记者｜范芊芊 每经编辑｜段炼 张喜威 王晓波 标题点睛： 当地时间9月29日，美国白宫公布了特朗普关于结束加沙冲突的计划。根据该计划，如果冲突双方同意，"战争将立即结束"。以色列军队将撤回到商定边 界，为释放人质做准备。在此期间，所有军事行动，包括空中和炮兵轰炸，都将暂停，战线将保持冻结状态，直到完全分阶段撤军的条件得到满足。协议 规定，在以色列公开接受此协议后的72小时内，所有的人质，无论生死，都将被归还。接受本协议后，将立即向加沙地带提供全额援助。 协议还要求哈马斯和其他派别不在加沙的治理中发挥任何直接、间接或任何形式的作用。所有军事设施都将被摧毁，不得重建。美国将与阿拉伯和国际伙 伴合作，发展一支临时国际稳定部队（ISF），以稳定局势。协议还规定，以色列不会占领或吞并加沙。（央视新闻） 当地时间9月29日，卡塔尔外交部发 ...

Group 1 - The U.S. stock market saw all three major indices rise, with the Dow Jones up 0.15%, Nasdaq up 0.48%, and S&P 500 up 0.26%. Major tech stocks had mixed results, with Nvidia rising over 2% and Amazon over 1%, while Broadcom fell nearly 2% [4] - The Nasdaq China Golden Dragon Index increased by 2.03%, indicating a general rise in Chinese concept stocks, with Bilibili, Alibaba, and New Oriental all gaining over 4% [4] - Spot gold prices reached a new high, closing at $3832.94 per ounce, while COMEX gold futures rose by 1.42% to $3862.90 per ounce [4] Group 2 - The Chinese Ministry of Industry and Information Technology, along with five other departments, released a plan for the mechanical industry aimed at maintaining steady growth from 2025 to 2026, targeting an average annual revenue growth rate of around 3.5% and aiming for total revenue to exceed 10 trillion yuan [8] - The National Development and Reform Commission indicated that while the economy is generally stable, there are still challenges ahead, and they will continue to implement macro policies as needed [10] - The State Taxation Administration announced that platform enterprises cannot transfer their tax obligations to workers, emphasizing compliance with tax regulations [11] Group 3 - China Mobile received a satellite mobile communication business operating license, allowing it to expand its service offerings in emergency communications and remote area connectivity [16] - Huawei appointed Yu Chengdong as the head of its Investment Review Board, indicating a strategic focus on AI development [18] - Alibaba's Tongyi AI models dominated the global open-source rankings, showcasing the company's leadership in AI technology [20] Group 4 - Meituan launched nighttime drone delivery services in Shenzhen, enhancing delivery efficiency and expanding its service range [26] - Seres completed the payment for a 10% stake in Shenzhen Yingwang Intelligent Technology from Huawei, indicating a deepening collaboration in smart technology [28] - AstraZeneca plans to list its shares on the New York Stock Exchange while retaining its headquarters in the UK, aiming to enhance its international presence [30]

DeepSeek has gone viral. Chinese AI lab DeepSeek broke into the mainstream consciousness this week after its chatbot app rose to the top of the Apple https://t.co/0GO2Z8ClH9 ...

Researchers at DeepSeek released a new experimental model designed to have dramatically lower inference costs when used in long-context operations. https://t.co/EidKhbISYw ...

Hey Mike, is pitching its newest model as a meaningful step toward real world AI agents. So systems that can operate independently and carry out tasks without constant human input. These autonomous assistants are the latest front in the AI model race with OpenAI just this hour announcing that it's switching on instant checkout in chat GBT so that users can buy directly in chat from Etsy and soon from Shopify.The move that's powered by its own agent tech have sent shares of Etsy and Shopify higher. Now, Inst ...

Core Insights - DeepSeek has released the V3.2-Exp model, which introduces a Sparse Attention mechanism aimed at optimizing training and inference efficiency for long texts [1] - The official app, web version, and mini-program have all been updated to DeepSeek-V3.2-Exp, and the API has seen a significant price reduction [1] - Under the new pricing policy, the cost for developers to access the DeepSeek API will decrease by over 50% [1] - The performance of DeepSeek-V3.2-Exp on various public evaluation datasets is comparable to that of V3.1-Terminus [1]

Core Insights - The new pricing policy will reduce the cost for developers using the DeepSeek API by over 50% [2][3] - The release of the DeepSeek-V3.2-Exp model on September 29, 2025, introduces the DeepSeek Sparse Attention mechanism, enhancing training and inference efficiency for long texts [2] - The V3.2-Exp model maintains performance levels comparable to the previous V3.1-Terminus model across various benchmarks [2][3] Performance Comparison - In the MMLU-Pro benchmark, DeepSeek-V3.1-Terminus scored 85.0, while V3.2-Exp maintained the same score [3] - For the BrowseComp search benchmark, V3.2-Exp improved to 40.1 from 38.5 in V3.1-Terminus [3] - The Codeforces-Div1 benchmark saw an increase from 2046 in V3.1-Terminus to 2121 in V3.2-Exp [3] Accessibility and Development - The V3.2-Exp model has been made open-source on Huggingface and Modao platforms, allowing users to access and develop further [5] - The updated version is available on the official app, web, and mini-programs [2][3]

Core Insights - DeepSeek has officially released and open-sourced the DeepSeek-V3.2-Exp model on the Hugging Face platform, marking a significant step towards the next generation architecture [1] - The new model introduces the DeepSeek Sparse Attention (DSA) mechanism, which aims to optimize training and inference efficiency for long texts while reducing computational resource consumption [1] - The model supports a maximum context length of 160K, with successful adaptations completed by Huawei and Cambricon [1] Technical Breakthroughs - The DeepSeek Sparse Attention (DSA) mechanism achieves fine-grained sparse attention, significantly enhancing training and inference efficiency for long text scenarios without compromising output quality [1][3] - The training settings for DeepSeek-V3.2-Exp were strictly aligned with the previous version, V3.1-Terminus, showing comparable performance across major public evaluation datasets [3] Benchmark Performance - Performance comparison between DeepSeek-V3.1-Terminus and DeepSeek-V3.2-Exp across various benchmarks shows: - MMLU-Pro: 85.0 (both versions) - GPQA-Diamond: 80.7 (V3.1) vs 79.9 (V3.2) - Humanity's Last Exam: 21.7 (V3.1) vs 19.8 (V3.2) - BrowseComp: 38.5 (V3.1) vs 40.1 (V3.2) - SimpleQA: 96.8 (V3.1) vs 97.1 (V3.2) - Codeforces-Div1: 2046 (V3.1) vs 2121 (V3.2) - AIME 2025: 88.4 (V3.1) vs 89.3 (V3.2) [4] Cost Reduction - The introduction of the new model has led to a significant reduction in API service costs, with a price drop of over 50%, effective immediately [4] Open Source and Community Support - DeepSeek has fully open-sourced the DeepSeek-V3.2-Exp model on Hugging Face and ModelScope, along with related research papers [6] - The company has retained API access for the V3.1-Terminus version for comparison purposes until October 15, 2025, with pricing aligned to V3.2-Exp [6] - To support community research, DeepSeek has also open-sourced GPU operators designed for the new model, recommending the use of the TileLang version for ease of debugging and rapid iteration [6] Industry Collaboration - Cambricon has announced the completion of adaptation for the new model and has open-sourced the vLLM-MLU inference engine source code, allowing developers to experience the new model's features on their hardware platform [6][7]

Core Insights - DeepSeek has released an experimental version of its model, DeepSeek-V3.2-Exp, which introduces Sparse Attention for improved training and inference efficiency on long texts [1] Group 1: Model Development - The new version, V3.2-Exp, is a step towards a next-generation architecture, building on the previous V3.1-Terminus [1] - The Sparse Attention mechanism is aimed at optimizing the model's performance for long text processing [1] Group 2: Pricing and Accessibility - The API pricing has been significantly reduced, with costs now at 0.2 yuan per million tokens for cache hits, 2 yuan for cache misses, and 3 yuan for output [1] - This pricing represents a reduction of over 50% compared to previous costs for developers using the DeepSeek API [1]