[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-cognition-kimi-k2-7-swe-1-7":3,"topics-all":36,"news-related-232841a4-204a-4530-a8f4-6bfc25ef16d8":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"232841a4-204a-4530-a8f4-6bfc25ef16d8","Cognition 把 Kimi K2.7 训成 Devin 级：SWE-1.7 击穿\"后训练天花板\"假设","Cognition 昨日发布 SWE-1.7——在已做过深度 RL 后训练的 Kimi K2.7 基座上,再用一轮大规模异步强化学习,把开源编程模型推到与 GPT-5.5、Opus 4.8 同档水准,成本只有闭源对手的一成。基准上,SWE-1.7 在 FrontierCode 1.1 Main 拿下 42.3%,比 Kimi K2.7 Code 高出 12 个百分点,几乎追平 GPT-5.5,与 Opus 4.8 仅差 4 点;每任务约 1.97 美元,通过 Cerebras 在 Devin 上以 1000 TPS 实时提供。训练工程是亮点。Cognition 用四块核心创新让\"后训练还能再涨 12 点\"成立:**top-p 采样重放**压平长 RL 的熵坍缩;**跨三洲多集群 RL**让 1T 参数模型跨大陆更新只需 1-2 分钟;**self-compaction + 交替长度惩罚**使单次 rollout 拉到 6 小时;**高质量验证器数据管线**过滤低信号样本,作弊一律奖励 0。SWE-1.7 打脸了\"基座 RL 已榨干\"的悲观叙事——同一个 K2.7 在 Cognition 手里再涨 12 点,说明 RL 后训练天花板远未触及;\"跨洲分布式 RL + Cerebras 1000 TPS 推理\"的工程组合,也为中小团队\"训出前沿编程模型\"提供了一条可复制样本。","https:\u002F\u002Fcognition.com\u002Fblog\u002Fswe-1-7","a6ca32cd-9c26-47a3-80c5-3cd215b56251",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b1853a5a-d940-42b7-94f9-0488ee3f2cf7","new-model",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"b608e682-8cb0-4b26-b2c0-bdcfbd3302fb","en","Cognition trains Kimi K2.7 to Devin level with SWE-1.7","Cognition released SWE-1.7 yesterday — using a large-scale asynchronous reinforcement learning round on the already-deeply-RL-post-trained Kimi K2.7 base, pushing the open-source programming model to the same tier as GPT-5.5 and Opus 4.8, with cost only one-tenth of closed-source competitors. On benchmarks, SWE-1.7 takes 42.3% on FrontierCode 1.1 Main, 12 percentage points above Kimi K2.7 Code, almost catching GPT-5.5, and only 4 points behind Opus 4.8; about $1.97 per task, served in real-time on Devin at 1000 TPS through Cerebras. The training engineering is the highlight. Cognition uses four core innovations to make \"post-training can still gain 12 points\" hold: **top-p sampling replay** flattens the entropy collapse from long RL; **cross-three-continent multi-cluster RL** lets a 1T-parameter model update across continents in just 1-2 minutes; **self-compaction + alternating length penalty** stretches a single rollout to 6 hours; **high-quality verifier data pipeline** filters low-signal samples, with cheats always rewarded 0. SWE-1.7 slaps the pessimistic narrative of \"base RL has been squeezed dry\" — the same K2.7 in Cognition's hands gains another 12 points, showing the RL post-training ceiling is far from reached; the \"cross-continent distributed RL + Cerebras 1000 TPS inference\" engineering combination also provides a replicable sample for small and medium teams to \"train a frontier programming model\".","cognition-kimi-k2-7-swe-1-7","2026-07-09T08:08:55Z","2026-07-09T08:08:55.322444Z","2026-08-19T02:08:40.142862Z",true,"agent",232,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"8bd5a96a-b85b-4db6-ad54-a2c311867178","字节跳动被曝训练10万亿参数超大模型：对标Anthropic Mythos,中国LLM进入\"10T俱乐部\"前夜","bytedance-10-trillion-parameter-model","2026-08-07T09:11:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"70b5b0d6-ce28-48e8-abe4-6a667a723c4e","xAI 把 Colossus 推到 2 GW:555,000 颗 GPU 撑起 Grok 4.6\u002F4.7 的万亿参数竞速","xai-colossus-2gw-grok-4-6-7-compute","2026-07-31T04:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"f4af1e4e-98a3-4810-831c-699ffe31ae73","马斯克公布 Grok 4.6\u002F4.7 路线图：1.5T\u002F2.1T 参数，SFT+RL 升级，8 月 7 日发行","grok-4-6-4-7-roadmap-1-5t-2-1t","2026-07-30T08:45:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"dea38861-2618-4468-9bad-a18eea96a818","Base44 Base 1：年入 1.5 亿美元的 vibe-coding 平台，终于把自己的 LLM 训出来","base44-base-1-vibe-coding-llm-launch","2026-07-29T06:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"fca4b7a0-4dd9-4475-bd64-ebd7667c7f58","MirrorCode 把长程编程拖进可测量区间：Opus 4.7 重写 6 万行 Pkl，AI 编码能力一年翻倍","mirrorcode-long-horizon-opus-pkl-56pct","2026-06-28T02:03:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"23dffa70-3b3e-452d-9730-a9c0074556ee","TMax 把「极简 RL」做成终端 Agent 工程范本:UW×Ai2 用 9B 模型跑出 27.2%,开源 14,600 训练环境","tmax-uw-ai2-terminal-agent-9b-27pct","2026-06-25T06:00:00+00:00"]