Introducing Nano Banana 2 Lite 🍌 and Gemini Omni Flash 🔮, our new generative media models in the Gemini API and AI Stu...
Introducing Nano Banana 2 Lite 🍌 and Gemini Omni Flash 🔮, our new generative media models in the Gemini API and AI Stu...
美团发布开源编码模型LongCat-2.0,采用1.6T参数MoE架构(活跃参数33B-56B),支持1M tokens上下文窗口。该模型在5万块国产芯片上从头训练,使用华为HCCL通信库,验证国内算力集群可胜任大模型预训练。已开源至longcat[.]ai和OpenRouter,调用量全球前三。与DeepSeek-V4-pro仅推理使用国产硬件不同,LongCat-2.0预训练和推理均依赖国产芯片。
关联讨论 4 条Hacker News 热门(buzzing.cc 中文翻译)IT之家(RSS)公众号:卡尔的AI沃茨公众号:龙猫LongCat(美团)🚨 SCOOP: Claude Sonnet 5 is releasing later today: - Knowledge cutoff January 2026 - Launching at $2/$10 per Mtok promo...
Introducing Nano Banana 2 Lite 🍌 and Gemini Omni Flash 🔮, our new generative media models in the Gemini API and AI Stu...
We're shipping 2 major releases: 🔘 Nano Banana 2 Lite: our fastest and cheapest Gemini Image model 🔘 Gemini Omni Flash...
GOOGLE 🔥: A new Nano Banana 2 Lite image generation model is on the horizon! > "Introducing Nano Banana 2 Lite for ultr...
美团发布基座推理模型LongCat-2.0(v2),采用MoE架构,总参1.6T,活跃约48B,支持1M上下文。专为智能体编程设计,引入LongCat Sparse Attention、Zero-Compute Experts及MOPD任务路由。基准测试中SWE-bench Pro达59.5(超GPT-5.5的58.6),多项Agent评测领先。模型已在OpenRouter上线,技术博客公开。美团强调全栈自研与低成本,v2基于ASIC训练。
Introducing LongCat-2.0 🐱 1.6T parameters · MoE with ~48B active · 1M context The full model behind Owl Alpha on @OpenR...
GOOGLE 🔥: A new Nano Banana 2 Lite image generation model is on the horizon! > "Introducing Nano Banana 2 Lite for ultr...
Woke up to sonnet 5 in the model selector We cannot use it yet unfortunately
Introducing LongCat-2.0 🐱 1.6T parameters · MoE with ~48B active · 1M context The full model behind Owl Alpha on @OpenR...
🚨 SCOOP: Claude Sonnet 5 is releasing later today: - Knowledge cutoff January 2026 - Launching at $2/$10 per Mtok promo...
⚡️🍌🔮
Introducing LongCat-2.0 🐱 1.6T parameters · MoE with ~48B active · 1M context The full model behind Owl Alpha on @OpenR...
关联讨论 4 条Hacker News 热门(buzzing.cc 中文翻译)IT之家(RSS)公众号:卡尔的AI沃茨公众号:龙猫LongCat(美团)美团 LongCat 推出旗舰模型 LongCat-2.0,采用 1.6T 参数 MoE 架构(约 48B 活跃参数),原生支持 1M 上下文窗口。定价为 Input Cache $0.015/1M tokens、Input $0.75/1M tokens、Output $2.95/1M tokens。模型专为 Agentic Coding 设计,包含三大技术:LSA 稀疏注意力实现高效 1M 扩展;Zero-Compute Experts 动态激活 33B–56B 参数/token,无算力浪费;MOPD 将专家分为 Agent / Reasoning / Interaction 三组,按任务门控路由。在 SWE-bench Pro 上取得 59.5 分,性能接近主流闭源模型。现已上线 SiliconFlow Day 0 服务。
同一事件,精选展示《美团 LongCat-2.0 正式发布:国产算力集群训练的万亿参数大模型》美团发布LongCat-2.0,1.6T参数MoE架构,激活参数~48B,上下文窗口1M(最大输出128K),使用5-6万张国产加速卡训练,训练推理全程零英伟达依赖。核心技术包括N-gram Embedding降低路由通信开销、稀疏注意力+跨层索引支撑长上下文、自研底层算子弥补国产芯片生态。定位Agent+Coding优先,非通用对话。Benchmark:Terminal-Bench 2.1 70.8,SWE-bench Pro 59.5(超GPT-5.5的58.6),SWE-bench Multilingual 77.3,FORTE 73.2等。与DeepSeek V4参数规模相近但路径不同:DeepSeek开源+双栈,LongCat强调全链路国产化。
Introducing LongCat-2.0 🐱 1.6T parameters · MoE with ~48B active · 1M context The full model behind Owl Alpha on @OpenR...
关联讨论 4 条Hacker News 热门(buzzing.cc 中文翻译)IT之家(RSS)公众号:卡尔的AI沃茨公众号:龙猫LongCat(美团)Imagine a lightweight AI that can read images AND chat with you. That's MiniCPM-V-4.6. It's a multimodal model that unde...
美团 LongCat 推出 LongCat-2.0,基于 MoE 架构,总参数 1.6T,激活参数约 48B,支持 1M 上下文。模型专为智能体编码设计,包含 LongCat 稀疏注意力(LSA)、零计算专家(33B–56B 动态激活)及 MOPD(三组任务路由专家)。基准测试:Terminal-Bench 2.1 达 70.8,SWE-bench Pro 59.5(超 GPT-5.5 的 58.6),SWE-bench Multilingual 77.3,FORTE 73.2,RWSearch 78.8,BrowseComp 79.9。目前已通过 OpenRouter 的 Owl Alpha 开放使用。
关联讨论 4 条Hacker News 热门(buzzing.cc 中文翻译)IT之家(RSS)公众号:卡尔的AI沃茨公众号:龙猫LongCat(美团)OpenAI 推出 GPT-5.6 模型套件的 limited preview,包含旗舰模型 Sol、中等模型 Terra 和快速廉价的日常模型 Luna。根据 GPT-5.6 Preview System Card,Sol 在内部编码测试中采取 severity-3 agent 动作的可能性比 GPT-5.5 高出近 10 倍。
Meta发布Brain2Qwerty v2,一种非侵入式脑机接口系统,能从实时脑信号解码完整自然句子,单词准确率达61%。系统基于约22000个句子训练,9名志愿者每人使用MEG记录10小时。相比此前非侵入方法8%的准确率大幅提升,最佳参与者达78%,超半数解码句子仅错一个词或更少。该端到端管线能实时将原始脑信号解码为单词和语义。但研究仍在受控实验室阶段:参与者样本小、依赖MEG硬件、数据来自主动打字、结果由公司报告,尚未成为临床通信设备。Meta已开源训练代码,BCBL发布v1数据集。
We're sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2. Building on ...
Some of you guessed right. 👀 Owl Alpha on @OpenRouter - that's us. Since going live, it has reached Top 3 globally by d...
关联讨论 4 条Hacker News 热门(buzzing.cc 中文翻译)IT之家(RSS)公众号:卡尔的AI沃茨公众号:龙猫LongCat(美团)商汤推出 SenseNova-U1-8B-MoT-Infographic 模型,能够生成工作室级别的高密度信息图,此前这类工作流程缓慢且昂贵。YouTuber CAPITAL R 制作了演示视频,模型已在 HuggingFace 上线,GitHub 页面展示示例图片,并开放 Discord 社区。
In addition to their excellent and unique training data, the Cursor team is also making major engineering contributions ...
OpenAI的GPT 5.6 Sol正在灰度测试,可通过Juice测试Prompt验证:选择gpt-5.5并设置推理为xhigh,运行Juice提示,若返回128则说明被灰度到GPT 5.6 Sol,否则仍是GPT 5.5(返回768)。社区报告Codex可能悄悄将部分gpt-5.5 xhigh会话路由至GPT 5.6 Sol,建议在Codex App/CLI中尝试验证。宝玉(@dotey)实测结果仍为768,说明未被灰度覆盖。
Community report: Codex may be quietly routing some gpt-5.5 xhigh sessions to gpt-5.6-sol. Try it in Codex App/CLI: sele...
To be clear, I'm not saying the Grok v9 foundation model will be mind-blowingly better than anything, but it will be a s...
马斯克宣布Grok 4.5基于1.5T V9基础模型,并在补充训练中加入Cursor数据,现已于SpaceX和Tesla进入私人测试。早期评估显示其性能接近甚至可能超过Opus。RL持续显著优化模型,Grok Build工具每日改进。此外,SpaceX今年将每月发布完全从零训练的新模型。
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta...
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta...
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta...
OpenAI 发布 GPT-5.6 系列,包括旗舰 Sol、均衡 Terra 和速度型 Luna。Sol 在 Terminal-Bench 2.1 得分为 88.8%(Ultra 模式 91.9%),领先 GPT-5.5 的 88.0% 和 Claude Mythos 5 的 84.3%;GeneBench v1 以更少输出 tokens 获更强结果,ExploitBench 接近此前 Mythos 但仅用约 1/3 输出 tokens。价格:Sol 输入 $5/百万 tokens、输出 $30,缓存读取九折。发布前 OpenAI 向美国政府展示能力,按政府要求先以有限预览上线,首批约 20 家合作伙伴可访问。