大模型升级
Search documents
DeepSeek网页版大升级!随后宕机11小时崩上热搜,新模型真的来了
猿大侠· 2026-03-31 04:40
Core Viewpoint - DeepSeek experienced a significant service outage lasting over 8 hours, which has been interpreted as a precursor to a model upgrade rather than a typical service interruption [1][2]. Group 1: Service Interruption and Model Upgrade - The outage was preceded by user reports indicating noticeable changes in the DeepSeek web version, suggesting a substantial enhancement in model capabilities [4]. - For instance, the model's performance in generating an SVG image of a pelican riding a bicycle improved significantly from the previous week [5]. - DeepSeek has a history of silently upgrading its models without prior announcements [9]. Group 2: Model Version and Knowledge Cutoff - The model's identity has been updated, with the March 29 version consistently introducing itself as DeepSeek-V3, contrasting with the previous vague self-description [11]. - The knowledge cutoff date has also changed, with the model now aware of the results of the 2025 U.S. elections but not the events of February 2026, indicating a potential knowledge cutoff around January 2026 [12]. Group 3: Performance Improvements - The model demonstrated a significant improvement in generating code for front-end pages on March 29 [15]. - There is speculation about whether this is a fine-tuned version of V3 or a direct upgrade to V4, as the official stance from DeepSeek remains unclear [17]. Group 4: Current Status and Future Prospects - The DeepSeek web version has resumed service, although minor issues persist, such as the model stopping output after deep thinking mode without displaying answers in the main text [19]. - Without deep thinking mode, the model appears to revert to an older version based on its self-introduction [21]. - The recent recruitment of 17 positions in the Agent direction suggests that DeepSeek may be preparing for significant developments after a period of silence [22].
DeepSeek网页版大升级!随后宕机11小时崩上热搜,新模型真的来了
量子位· 2026-03-30 02:35
Core Viewpoint - DeepSeek experienced a significant service interruption lasting over 8 hours, which users interpreted as a potential model upgrade rather than a typical outage [1][2]. Group 1: Service Interruption and Model Upgrade - The service disruption was reported by users who noted changes in the DeepSeek web version, indicating a substantial enhancement in model capabilities [4]. - For instance, the model's performance in generating SVG images, such as a pelican riding a bicycle, showed marked improvement on March 29 compared to the previous week [5]. - DeepSeek has a history of silent model upgrades without prior announcements, suggesting this may not be an isolated incident [8]. Group 2: Model Versioning and Knowledge Cutoff - The updated version, identified as DeepSeek-V3, provides a more stable self-introduction compared to the previous version, which lacked clarity on its version number [10]. - The knowledge cutoff date appears to have changed, with the model now aware of U.S. election results up to 2025 but not events from February 2026, indicating a possible cutoff around January 2026 [11]. Group 3: Performance and Future Developments - The model's ability to generate code for front-end pages has significantly improved as of March 29 [14]. - Despite the service restoration, some issues remain, such as the model ceasing output after deep thinking mode, which does not display answers in the main text [18]. - The company has recently opened 17 positions related to agent development, hinting at significant upcoming advancements [21].
还在等DeepSeek R2?刚刚,DeepSeek R1模型小版本试升级已完成!优化了这些方面
Mei Ri Jing Ji Xin Wen· 2025-05-28 13:03
Core Viewpoint - DeepSeek has announced the completion of a minor version upgrade for its R1 model, inviting users to test the new features on its official website, app, and mini-programs while maintaining existing API interfaces and usage methods [1]. Group 1: Upgrade Features - The upgrade focuses on several key areas: 1. Response quality optimization, enhancing accuracy in complex reasoning and multi-step calculations, as well as improving coherence and clarity in long text understanding and generation, and reliability in specialized outputs like mathematics and programming [2]. 2. A slight improvement in response speed, with a 10% to 20% reduction in latency, particularly when processing long text inputs across web, app, and API interfaces [2][4]. 3. Enhanced dialogue stability, with improved context memory, especially in long conversations, supporting up to 128K context and reducing instances of "forgetting settings" or "going off track" [4]. 4. API and interface compatibility remains stable, with no changes to API calling methods, parameters, or return structures, allowing users to seamlessly use the new version without adjustments [5]. Group 2: Upgrade Process - The upgrade is termed a "trial upgrade" due to: 1. It being a "gray release," where a portion of users will experience the upgrade first [6]. 2. The company will collect feedback to ensure stability before a full rollout [6]. 3. Users of the official app, website, or mini-program may already be using the upgraded version in "Deep Thinking" mode [6]. Group 3: Future Developments - There is ongoing speculation regarding the release of the DeepSeek R2 model, with the company previously denying rumors about its launch on March 17 [6].