跳到正文
10月28日周三
11月21日周六
  1. OpenAI News42

    Hebbia 测试 GPT-6 Astra 在金融研究中的表现

    Hebbia 测试 OpenAI 的 GPT-6 Astra,在金融研究任务中取得显著优势。该模型在引用所需源文件类型方面比最佳替代方案高出 13.6 个百分点,且 99.5% 的引用摘录与问题相关。在演示文稿生成方面,GPT-6 Astra 满足 97.9% 的适用要求,领先竞争对手 14.4 个百分点,实现了指令遵循与呈现质量的双重领先。

11月20日周五
  1. The Verge · AI42

    Apple 招聘工程师构建自有 AI 智能体

    Apple 正招聘工程师开发自有 AI 智能体,旨在跨应用执行浏览网页、编写代码及调用服务等多步骤任务。该计划将增强 Siri 的自主操作能力,并配合 iOS 27.3 更新允许开发者在更多应用中集成 Siri。此举使 Apple 加入与 Meta Muse、OpenAI Dots 等产品的竞争行列。

11月14日周六
  1. 钛媒体42

    字节的2029与腾讯的“一半人”:AI时代如何争取时间

    张一鸣在字节AI团队内部会议上称,字节真正追平OpenAI和Anthropic等前沿最早要到2029年或2030年,并计划几周内推出对标Meta Muse的AI助手。马化腾在腾讯28周年讲话中提出一半人力守核心业务提效、一半投入AI Native探索,该说法被传为“裁员50%”后遭公关总监张军否认。字节将2026年AI基础设施预算上调至2000亿元以上,腾讯上半年资本开支847.2亿元。

11月20日周五
11月18日周三
  1. MarkTechPost38

    澄清 AI 数据保留误区:两种标准 API 设计解析

    OpenAI、Amazon Bedrock、Google Gemini 和 Fireworks AI 等平台的 API 遵循两种标准数据生命周期设计。第一种为“上传-使用-删除”模式,如 Batch API,数据由客户管理并存储至删除或过期。第二种为“发送-计算-返回”的临时处理模式,适用于 Chat Completions 等推理接口,默认不持久化应用状态。

11月11日周三
  1. 机器之心38

    Anthropic 发布 Claude Fable 5.2,能玩魔方、拧瓶盖、操控机器人画画

    Anthropic 发布新模型 Claude Fable 5.2,除编程与推理升级外,还能玩魔方、拧瓶盖并操控机器人画画。此前 OpenAI 的 Astra 已在机器人操作任务上展示过类似能力,有研究认为这类表现可能依赖专门的机器人数据训练。简智同期推出数据手套 Dex 2.0 和第一视角头戴设备 DAS Ego 2.0,用于采集手部运动与触觉等多模态人类数据。

11月12日周四
11月13日周五
  1. IT之家88

    Anthropic IPO 认购需求约为 6 倍,拟募资最高 644 亿美元、目标估值约 1.4 万亿美元

    据彭博社报道,Anthropic 的 IPO 认购订单已达到可供认购股份数量的约 6 倍。公司及其联合创始人计划通过 IPO 募资最高 644 亿美元,并通过同步私募配售募资 260 亿美元,两项合计最高 904 亿美元,目标完全稀释后估值约 1.4 万亿美元。英伟达计划在私募配售中认购约 100 亿美元,承销银行已告知部分投资者将优先考虑现有投资者。

    推荐理由:材料给出 Anthropic IPO 的认购倍数、募资规模与估值区间,可据此判断市场对头部模型公司的资金热度。

11月11日周三
  1. 智东西88

    Anthropic启动IPO路演,英伟达与谷歌各投100亿美元

    Anthropic于11月10日晚启动IPO路演,发行价区间为每股70至78美元,计划发行8.25亿股,按上限计算最高募资643.5亿美元,完全稀释估值最高约1.405万亿美元。

    推荐理由:材料把IPO发行区间、七家科技公司私募认购和最新年化收入放在一起,可看清Anthropic上市前的资金与营收结构。

11月10日周二
  1. IT之家88

    Anthropic IPO 每股定价区间 70-78 美元,最高估值达 1.405 万亿美元

    Anthropic 更新上市申请文件,计划通过 IPO 发行 8.25 亿股,发行价区间为每股 70 至 78 美元,按上限计算最多募资 643.5 亿美元,完全摊薄估值最高 1.405 万亿美元。

    推荐理由:材料给出 IPO 定价区间、募资规模与七家科技公司的私募认购安排,可对照此前估值目标看市场预期的变化。

  2. OpenRouter85

    Anthropic 的 Claude Fable 5.2 已在 OpenRouter 上线,支持 1M 上下文,定价为输入 $8/M、输出 $40/M。OpenRouter 引用的信息称,该模型在多数评测上比 Opus 5.5 更强,输出速度比 Fable 5.1 快至多 30%;Anthropic 公布的数据为 FrontierSWE v2 69.9%、OSWorld 2.1 89.3%、Humanity's Last Exam(带工具)74.2%。

    引用Models on OpenRouter@ModelsOnOR

    Claude Fable 5.2 from @AnthropicAI is live on OpenRouter Anthropic's model for the most ambitious technical work, a step above Opus 5.5 on most of its evals. Up to 30% faster output than Fable 5.1, now at a lower price: $8/M input, $40/M output, 1M context. Anthropic's numbers: 69.9% on FrontierSWE v2, 89.3% on OSWorld 2.1 and 74.2% on Humanity's Last Exam with tools. See more: http://openrouter.ai/anthropic/claude-fable-5.2

    推荐理由:读者可据此了解 Claude Fable 5.2 在 OpenRouter 上的定价、上下文长度与 Anthropic 公布的评测数据。

10月21日周三
8月13日周四
11月10日周二
  1. IT之家78

    黄仁勋承认婚外情后退出 ElevenLabs 峰会等公开活动

    英伟达 CEO 黄仁勋已退出原定本周三在纽约举行的 ElevenLabs 峰会,ElevenLabs 于周一证实该消息,黄仁勋还取消了包括 Anthropic CEO 论坛在内的其他几场公开演讲安排。此事发生在他上周末承认婚外情之后,英伟达已就此展开调查,重点审查他与该女性的关系。黄仁勋在内部备忘录中向家人和员工道歉,事件已引发外界对其能否继续担任英伟达 CEO 的关注。

    推荐理由:黄仁勋在承认婚外情后接连退出公开活动,英伟达已启动调查,事件牵动其CEO职位前景。

11月1日周日
  1. IT之家78

    微软、谷歌、Meta 等五大巨头承诺 AI 基建投入超 3 万亿美元

    据彭博社报道,Alphabet、Meta、微软、亚马逊和甲骨文承诺未来数年在数据中心等领域投入超过 3 万亿美元。其中 Alphabet 的合同承诺总额超过 1 万亿美元,接近一年前的八倍;Meta 未来支出总额超过 8000 亿美元,近一半用于尚未生效的数据中心租约,最长分 30 年支付。

    推荐理由:原文列出五家巨头的租赁与采购承诺拆分,读者可据此看清 AI 数据中心投入的规模与回收争议。

10月31日周六
7月27日周一
  1. Kimi.ai78

    Kimi K3 已在 Nebius Token Factory 上线,Nebius 是 Kimi 的 Day 0 发布合作伙伴。据 Nebius 介绍,K3 面向长程编码、知识工作与推理,原生支持视觉、上下文最高 1M tokens,Artificial Analysis 智能指数得分为 57,开发者可通过兼容 OpenAI 的 API 和 console 接入。

    引用Nebius Token Factory@nebiustf

    Kimi K3 is now available on Token Factory. We’re excited to announce that Nebius Token Factory is an official Day 0 partner for @Kimi_Moonshot's Kimi K3. Kimi K3 is the first open-weight model to reach frontier-level performance, a major step forward for open models. It is built for long-horizon coding, knowledge work and reasoning, with native vision and up to 1M tokens of context. Artificial Analysis scores it at 57 on its Intelligence Index, just two points behind GPT-5.6 Sol (max). That puts Kimi K3 at the top of the open-weight field and firmly among today’s frontier models. Developers can access K3 through Token Factory’s OpenAI-compatible API and console today. Give K3 the hard problem. Build with Kimi K3: https://tokenfactory.nebius.com/?modals=endpoint-details&model-id=moonshotai/Kimi-K3

    推荐理由:Kimi K3 通过 Nebius Token Factory 上线,开发者可据此了解接入方式与模型定位。

10月20日周二
11月21日周六
  1. Kling AI72

    快手可灵发布 Kling Image 4.0,支持原生 8K、更逼真的画面、一致的图像组和高级文本渲染。据其引用的 ArtificialAnlys 评测,该模型在文生图排名第 7、图像编辑排名第 6,2K 价格不到排名更高 OpenAI 模型的三分之一;相比 Kling Image 3.0 Omni,两个榜单的 Elo 提升均超过 225。黑五限时活动中,2K 低至 $0.022/张,原生 8K 低至 $0.033/张。

    引用Artificial Analysis@ArtificialAnlys

    Kling Image 4.0 ranks #7 in Text to Image and #6 in Image Editing, behind only OpenAI and Microsoft AI models It generates images up to 8K, and its 2K price is under a third of the price of every OpenAI model ranked above it. In Image Editing, it sits just behind GPT Image 2.5 Flare (max) at #5 and ranks in the top 5 for 5 of 7 editing actions. It is also a big step up from Kling Image 3.0 Omni, which ranked #86 in Text to Image and #67 in Image Editing before this launch. Kling Image 4.0 gains more than 225 Elo on both boards and improves on Kling Image 3.0 Omni in every editing action and Text to Image capability. Congratulations to @Kling_ai on the launch of Kling Image 4.0! See below for our detailed analysis in Image Editing and Text to Image 🧵

    推荐理由:官方给出 Kling Image 4.0 的原生 8K 与文本渲染能力,并附第三方榜单排名和限时价格,便于判断其定位。

11月14日周六
  1. meng shao71

    Sierra 联合创始人 Bret Taylor 和 Clay Bavor 宣布个人智能体协议 Poppy v1.0 正式发布,该协议由 Sierra 与 Meta 等合作方共同推动。

    引用Sierra@SierraPlatform

    Today we’re releasing Personal Agent Protocol (Poppy) v1.0. It’s on for everyone who uses Muse, @Meta's AI agent, enabled for every merchant on @Shopify, and backed by more than 100 design partners.

    推荐理由:原文梳理了 Poppy 1.0 的授权接入流程与 Shopify、Meta Muse 的落地情况,也点明规范仍是草案。

11月13日周五
  1. Andrew Curran88

    Will DePue 发布取证报告称,DeepSeek、Moonshot、MiniMax 和蚂蚁集团自 2026 年 4 月以来发布的六款模型,其公开权重中带有 NVIDIA Blackwell GPU 的初始化指纹,包括 DeepSeek-V4-Pro、Kimi K3 和 MiniMax-M3。他搜索超过六万亿种芯片布局与种子组合,并在真实硬件上复现:B300 逐位复现 DeepSeek-V4-Flash 的初始权重,租用的 GB200 复现 Kimi K3 的相关系数为 0.999995,而 H100 无法复现。他估计 Kimi K3 与 DeepSeek-V4-Pro 各使用 2000 至 3000 块 Blackwell GPU,并称结果已获 OpenAI 和 Anthropic 资深研究员独立验证;报告同时说明该指纹支持但未证明 Blackwell 也用于训练,也未确认芯片位置与获取方式。

    引用will depue@willdepue

    I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at http://unemploymentlabs.com/checkpoint-forensics. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

    同一件事,精选收录的是《前 OpenAI 研究员称在六个中国 AI 模型中检出 Nvidia Blackwell 硬件特征》

  2. 宝玉88

    Google 于 11 月 12 日发布 Gemini 4 Hydrogen,接替 Gemini 3.8 Flash,主打快和省,对标 Anthropic 的 Claude Sonnet 5.5 和 OpenAI 的 GPT-6.1 Terra。

    引用Sundar Pichai@sundarpichai

    Following Carbon + Argon, Gemini 4 Hydrogen brings cutting-edge intelligence to everyone. Built for speed and efficiency for everyday tasks, coding and agentic workflows. Compared to 3.8 Flash, it finishes tasks 2.5x faster. Also has 4 thinking levels, so you only wait and pay for the reasoning depth a task requires. Rolling out today to all tiers of @GoogleAIStudio, the Gemini API, @GeminiApp, Search and paid Google AI Pro and Ultra subscribers in Google @Antigravity.

    同一件事,精选收录的是《Google DeepMind 发布 Gemini 4 Hydrogen 模型》

10月26日周一
  1. 宝玉88

    Anthropic 于美国时间 10 月 26 日向 SEC 公开递交 IPO 招股书(S-1),计划在纳斯达克上市,股票代码 ANT,由高盛和摩根士丹利牵头承销,发行股数和定价尚未确定。

    引用Anthropic@AnthropicAI

    Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO. Read more: https://www.anthropic.com/news/s1-sec Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.

    同一件事,精选收录的是《Anthropic 就拟议 IPO 向 SEC 递交 S-1 注册声明》

10月23日周五
11月11日周三
11月3日周二
  1. Rohan Paul78

    Anthropic has committed $895.4B to computing and infrastructure, with $447.5B due before 2031. $161 B in Broadcom equipment leases, $121 B with Google, up to $120B with SpaceX and $110B with Amazon. With Amazon and Microsoft backlogged, Anthropic signed up GPU renters like Lambda and former crypto miners like TeraWulf, which already control land and power. To cut delays further, it buys Nvidia, AMD and Broadcom chips and cooling gear itself, so site operators can install equipment faster. Only $213.8B can be cancelled or hinges on conditions like partners securing financing, so most must be paid even if demand slows. It already warns it may need money sooner than expected, so staying at the frontier hinges on revenue outrunning those payments.

    引用Rohan Paul@rohanpaul_ai

    Anthropic 已承诺投入 8954 亿美元用于计算和基础设施,其中 4475 亿美元需在 2031 年前支付。 其中包括 1610 亿美元的 Broadcom 设备租赁、1210 亿美元与 Google 的合作、最高 1200 亿美元与 SpaceX 的合作,以及 1100 亿美元与 Amazon 的合作。 由于 Amazon 和 Microsoft 订单积压,Anthropic 签约了 Lambda 等 GPU 租赁商,以及 TeraWulf 等前加密货币矿企——后者已掌握土地和电力资源。 为进一步减少延误,它自行采购 Nvidia、AMD 和 Broadcom 的芯片及冷却设备,以便站点运营方更快安装设备。 其中仅 2138 亿美元可取消或取决于合作伙伴获得融资等条件,因此大部分款项即使需求放缓也必须支付。 它已警告称可能比预期更早需要资金,因此能否保持在最前沿,取决于收入能否跑赢这些付款。

    同一件事,精选收录的是《Anthropic 更新 IPO 招股书:三季度收入 169.56 亿美元,累计算力承诺约 8954 亿美元》

10月30日周五
10月20日周二
11月18日周三
  1. Turing Post (Ksenia Se)68

    斯坦福一项论文让 OpenAI、Anthropic、Google 和 Qwen 的 8 个模型回答 620 个真实临床问题,再匿名互评,发现每个评委都偏好自己的答案,平均排名高出 1.22 位,GPT-5.6 Sol 接近 3 位。作者指出,医疗 AI 基准常用 LLM 给 LLM 打分,评委偏好自身输出会让基准偏向该模型;即使去掉评分细则,这种偏好依然存在,公开模型名称也只让偏差缩小 0.094 位。作者建议使用跨模型家族的评委小组,并按评委分别报告结果。

    引用Rohan Paul@rohanpaul_ai

    New Stanford paper finds that all 8 LLMs tested rank their own medical answers higher than other models rank them, even when answers are anonymous. Medical AI benchmarks often use LLMs to grade other LLMs' answers. If a judge favors its own outputs, the benchmark tilts toward that model. They had 8 models from OpenAI, Anthropic, Google and Qwen answer 620 real clinician questions, then rank all 8 anonymous answers. They compared where each judge placed its own answer with where the other 7 judges placed that same answer. Every judge favored its own answer, by 1.22 rank positions on average and nearly 3 for GPT-5.6 Sol. The bias held without a rubric, and revealing model names cut it by only 0.094 positions. If you use LLM judges, use a panel from several model families and report results per judge instead of 1 pooled score.

11月2日周一
11月18日周三
  1. Andrew Curran65

    Vals 的评测结果显示,GPT-6.2 Astra 能在无人工输入的情况下完成持续一天的研究任务,在方法成熟的预训练任务上表现接近人类参考水平。Vals 称这一结果与 OpenAI 9 月所述已达成自动研究实习生目标相符,即系统可在人类指导下完成原本需要熟练研究者数天的明确研究任务。

    引用Vals AI@ValsAI

    OpenAI reported in September that it had reached its goal of an "automated research intern": a system that can carry out well-defined research tasks under human direction, including tasks that would take a skilled researcher a few days. Our results align with this classification. GPT-6.2 Astra completed day-long research tasks without human input and performed close to the human reference on pre-training tasks, where methods are well established.

    同一件事,精选收录的是《GPT-6.2 Astra 被指达到 OpenAI Preparedness Framework 自我改进 High 阈值》

11月16日周一
  1. Qwen62

    Arena 榜单显示,Qwen4-Max 在 Agent Arena 上较 Qwen3.8-Max 净提升 7.5%,整体排名从第 39 升至第 14,成为排名最高的中国模型,阿里也因此成为第 5 大实验室,仅次于 Anthropic、Google、OpenAI 和 Meta。该成绩基于全球社区 14.7k 条真实智能体会话轨迹;四项信号中用户满意度与可控性提升最大,Praise vs. Complaint 提升 13%、Confirmed Success 提升 9%、Steerability 提升 8%、Bash Reliability 提升 3%。中国模型中,腾讯 Hy4 以 +5.7% 排第 19,月之暗面 Kimi K3.1 Pro (Max) 以 +4.0% 排第 21。

    引用Arena.ai@arena

    Qwen4-Max by @Alibaba_Qwen is the #1 Chinese model on the leaderboard making Alibaba now the #1 Chinese lab and #5 lab overall, just behind Anthropic, Google, OpenAI and Meta. It debuts overall at #14 on Agent Arena with +7.5% net improvement across 14.7k real-world agentic sessions from our global community, real traces on real tasks. This is a major leap over Qwen3.8-Max - Rank: #39 → #14 overall The Chinese frontier on Agent Arena: - #14 Qwen4-Max (Alibaba): +7.5% - #19 Hy4 (Tencent): +5.7% - #21 Kimi K3.1 Pro (Max) (Moonshot): +4.0% - #26 MiMo-V2.6-Pro (Xiaomi): +1.5% - #30 DeepSeek-V4.1-Flash (Max) (DeepSeek): +1.1% Breaking down performance across signals: Qwen4-Max improves over Qwen3.8-Max across all four reported signals, with particularly large gains in user satisfaction and steerability. - #15 Praise vs. Complaint: +13% (best Chinese model) - #13 Confirmed Success: +9% (best Chinese model) - #19 Steerability: +8% - #29 Bash Reliability: +3% By category, it is a top-ranked Chinese model across all 3: - #12 in Work: +9% - #14 in Chat: +8% - #16 in Code: +8% Congrats to the @Alibaba_Qwen team on this release!

10月22日周四
  1. 钛媒体62

    利润腰斩的特斯拉,要靠合并SpaceX救场?

    特斯拉2026年三季度交付48.65万辆、营收280.5亿美元,营业利润同比腰斩至7.2亿美元,研发费用飙升69%至27.6亿美元。马斯克在SpaceX电话会上确认两家公司董事会已各自成立特别委员会评估合并,合并后估值约3.5万亿美元。文章还对比了比亚迪前三季度纯电累计163万辆领先特斯拉132.5万辆,以及小米YU7、SU7在细分市场反超Model Y和Model 3。

10月28日周三
  1. Aran Komatsuzaki60

    OpenBMB 发布 1.3B 开源 OCR 模型 MiniCPM-OCR 1.0,在 OCRBench v2 上从 47.4 提升到 54.7,CHAOS-Bench 从 21.28 提升到 29.78,吞吐为 PaddleOCR-VL-1.6 的 1.49 倍。

    引用OpenBMB@OpenBMB

    Your OCR says it’s done—but did it get everything right? 🤔 In high-stakes workflows, even a small error can be costly. 🚀 Meet the MiniCPM-OCR suite: an open-source solution to parse fast and guard the results. Two new models: 🔹 MiniCPM-OCR 1.0 — SOTA OCR performance and speed at 1B scale. Compact, accurate, and fast. 🔹 MiniCPM-OCR Jev — A 2B decision model that helps assess parsing accuracy, with performance approaching that of Qwen3.5-397B-A17B on JevOCRBench. ⚡ What’s new? Small model, SOTA results: MiniCPM-OCR 1.0 scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2. A sharp goalkeeper 🛡️: MiniCPM-OCR Jev is model-agnostic, checking text, formulas, and tables from different OCR parsers. It achieves 93.6 on JevOCRBench, outperforming JEV-35B-A3B (90.1) and approaching Qwen3.5-397B-A17B (93.8). Up to 128 pages per minute 🤯: With DSpark, MiniCPM-OCR runs at 1.49× the speed of PaddleOCR-VL-1.6 and 3.62× that of MinerU2.5 Pro. The SDK’s concurrency optimizations further boost throughput to as much as 128 pages per minute. Deploy where you work: Day-0 support for Arm, Houmo AI, Intel, and NVIDIA, plus 10 more chip vendors through FlagOS. Ready for your stack: vLLM, SGLang, llama.cpp, and Ollama, with quantized models, fine-tuning support, and Agent Skills. 🔗 GitHub: http://github.com/OpenBMB/MiniCPM-OCR 🤗 Hugging Face: MiniCPM-OCR 1.0:http://huggingface.co/openbmb/MiniCPM-OCR-1.0 MiniCPM-OCR Jev:http://huggingface.co/openbmb/MiniCPM-OCR-Jev 👉 Try the Demo:http://platform.modelbest.ai/console/playground/document-parse Use the API:http://platform.modelbest.ai/console/docs/models/document-parse

11月21日周六
  1. Lisan al Gaib58

    作者指出,METR Time Horizons 2.0 的倍增时间比 1.1 版本短得多。其引用的内容显示,TH 2.0 的倍增时间为 3.6 个月,接近 OpenAI 估计的 3.8 个月;GPT-6 Astra 为 40.3 小时,Opus 5.5 为 39.0 小时,但 95% 置信区间极宽,约从 11.9 小时到 110.2 天。配图还给出 TH1.1 倍增时间 3.07 个月、TH2(可验证)1.65 个月、MirrorCode 1.79 个月等估计。

    引用Lisan al Gaib@scaling01

    THEY ARE BACK! METR TIME HORIZONS 2.0 Doubling time of 3.6 months, similar to the 3.8 months estimate by OpenAI GPT-6 Astra: 40.3 hours Opus 5.5: 39.0 hours but with extremely wide 95% CI from around 11.9 hours to 110.2 days

11月19日周四
  1. SenseTime58

    商汤宣布 SenseNova U1 Pro 在 DesignArena 的图像生成和图像编辑两个榜单均首次上榜即位列第 6,使商汤在图像 AI 实验室中排名第二,仅次于 OpenAI。商汤还称 U1 Pro 进入图像偏好与价格的帕累托前沿,在较低生成成本下保持画质与美感。DesignArena 的引用内容提到,该模型在生成特定艺术风格或年代感图像时胜率较高,图像编辑则擅长按用户要求精确修改而不影响画面其他部分。

    引用DesignArena@DesignArena

    BREAKING: SenseNova U1 Pro by @sensetime_ai makes its debut as the #2 Image lab, just behind @OpenAI. For image generations, it wins most often on prompts that ask for specific art styles such as an “impressionist painting,” or an image set within a certain time period such as the 1920s art deco era. For image editing, its strength is in making precise, user-requested edits without creating unintended changes elsewhere in the image. Congrats to the @sensetime_ai team on achieving a strong result!