[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-gpt-6-astra-launch":3,"topics-all":38,"news-related-5bf8fa2d-258e-41d3-bfb6-5c2053433cfd":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"5bf8fa2d-258e-41d3-bfb6-5c2053433cfd","GPT-6 Astra 正式上线:8 月因安全被暂停的旗舰回来了","OpenAI 发布 GPT-6 Astra:1.05M token 上下文、10\u002F50 美元定价、五档推理强度。官方 ARC-AGI-3 成绩 99.9% 但依赖 adapter harness,独立复测 17-63%;安全基准领先,电脑使用与自动化是核心买点。","8 月中旬,OpenAI 第一次因安全问题搁置了最大规模的强化学习训练:Astra 在网络攻防能力上触发了「关键」阈值,最大 RL run 被暂停。两周多后,这个模型以正式形态上线——GPT-6 Astra 现已向 Trusted Access 项目内的企业客户开放,API 内名为 gpt-6-astra,同步登陆 Amazon Bedrock,ChatGPT Plus\u002FPro\u002FBusiness\u002FEnterprise 将在几天内跟进,Pro 及以上套餐还会获得 GPT-6 Astra Pro。\n\n## 规格与定价:1M 上下文,五档推理\n\n官方文档给出的核心参数:1,050,000 token 上下文窗口、128K 最大输出、知识截止 2026 年 4 月 30 日。API 定价每百万 token 输入 10 美元、缓存命中 1 美元、输出 50 美元;超过 272K 输入的请求按输入 2 倍、输出 1.5 倍计费。reasoning.effort 分 low\u002Fmedium\u002Fhigh\u002Fxhigh\u002Fmax 五档,Fast mode 提供 2.5 倍速度、收 2 倍价格。这个价位与 Claude Fable 5.1 的 10\u002F50 美元完全同档,对标意图明显。\n\n## 成绩单:电脑使用与自动化是最大买点\n\n官方评测表里,Astra 的优势集中在端到端长任务:AutomationBench 41.4%(GPT-5.6 Sol 为 18.1%,Claude Fable 5.1 为 31.4%),OSWorld 2.0 离线集 72.6%,ScreenSpot-Pro 92.7%,Agents' Last Exam 59.3%。长上下文 MRCR v2 8-needle 在 512K-1M 段拿到 96.3%(Sol 73.8%)。但有一个数字值得注意:Artificial Analysis 智能指数 v4.1.1 上 Astra 为 61.2,落后于 Claude Fable 5.1 的 65.7——通用智能并不是它的王牌。\n\n## 99.9% 的 ARC-AGI-3 要仔细读\n\n官方口径下 Astra 的 ARC-AGI-3 成绩是 99.9%,对比 Sol 的 7.8% 和 Claude Opus 5 的 30.2%。但这个数字来自 OpenAI 自己的 adapter harness;ARC Prize 的独立复测显示,标准无状态 harness 下成绩随推理档位在 17% 到 63% 之间浮动,完整复现一次「99.9% 配置」要花数万美元。两个口径都真实,但引用时必须带上条件。\n\n## 安全闭环:暂停的理由,成了发布的主角\n\n8 月的暂停理由是网络攻击能力触线,9 月的发布文里安全基准被放在显眼位置:ExploitGym 蜜罐测试中 Astra 的触发率为 0.0%(官方列出的对照模型为 48.2%),内部规避基准 0.00%,内部幻觉基准 4.2%(对照 12.2%)。独立实验室 Irregular 在 FrontierCyber 的 226 道挑战里测得 Astra 解出 86 题(Sol 为 34),其中包括浏览器和云数据库的 zero-day 漏洞——攻击能力确实到位,护栏数字因此才值得认真看。\n\n对按 token 采购的开发者,Astra 的 10\u002F50 美元买的不是聊天质量,而是「电脑使用 + 长上下文 + 安全余量」的组合;看到 99.9% 这类数字时,先问一句:harness 是谁的。\n\n参考: https:\u002F\u002Fopenai.com\u002Findex\u002Fgpt-6-astra\n规格与定价: https:\u002F\u002Fdevelopers.openai.com\u002Fapi\u002Fdocs\u002Fmodels\u002Fgpt-6-astra\n","https:\u002F\u002Fopenai.com\u002Findex\u002Fgpt-6-astra","15975962-b5fe-49e5-ae68-687ba6cb7015",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"baf131c1-687a-49f4-87f6-4dd87c1c692f","gpt",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":19,"name":20,"slug":20,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":22,"name":23,"slug":23,"description":14,"color":14},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"efd2b29b-2823-48c1-bc36-87894ce17d5d","en","GPT-6 Astra Ships With 1M Context After Safety Pause","GPT-6 Astra is live: 1.05M context, $10\u002F$50 pricing. Official ARC-AGI-3 99.9% needs an adapter harness; independent runs score 17-63%.","In mid-August, OpenAI halted its largest reinforcement learning run for safety reasons for the first time: Astra had crossed a \"critical\" threshold in cyber-offense capability. A little over two weeks later, the model shipped — GPT-6 Astra is now available to enterprises in the Trusted Access Program, exposed as gpt-6-astra in the API and on Amazon Bedrock, with ChatGPT Plus, Pro, Business and Enterprise access rolling out over the coming days; Pro and above also get GPT-6 Astra Pro.\n\n## Specs and pricing: 1M context, five reasoning tiers\n\nThe official docs list a 1,050,000-token context window, 128K max output, and an April 30, 2026 knowledge cutoff. API pricing is $10 per million input tokens, $1 cached, $50 output; requests beyond 272K input tokens are billed at 2x input and 1.5x output rates. reasoning.effort spans low, medium, high, xhigh and max, and Fast mode offers 2.5x speed at 2x price. That lands exactly on Claude Fable 5.1's $10\u002F$50 — the benchmarking intent is obvious.\n\n## The scorecard: computer use is the headline\n\nIn OpenAI's eval tables, Astra's edge concentrates in end-to-end long tasks: AutomationBench 41.4% (GPT-5.6 Sol 18.1%, Claude Fable 5.1 31.4%), OSWorld 2.0 offline set 72.6%, ScreenSpot-Pro 92.7%, Agents' Last Exam 59.3%. On long context, MRCR v2 8-needle reaches 96.3% in the 512K-1M band (Sol 73.8%). One number deserves attention: on the Artificial Analysis Intelligence Index v4.1.1, Astra scores 61.2 — behind Claude Fable 5.1's 65.7. General intelligence is not its trump card.\n\n## Read the 99.9% ARC-AGI-3 carefully\n\nOpenAI reports 99.9% on ARC-AGI-3, against 7.8% for Sol and 30.2% for Claude Opus 5. But that figure comes from OpenAI's own adapter harness; ARC Prize's independent runs show 17% to 63% on the standard stateless harness depending on reasoning tier, and a full reproduction of the 99.9% configuration costs tens of thousands of dollars. Both readings are real — just always attach the conditions.\n\n## The safety loop: the pause reason became the release headline\n\nAugust's pause was about cyber capability crossing a line; September's announcement puts safety benchmarks front and center: Astra triggers the ExploitGym honeypot at 0.0% (the comparison model listed by OpenAI: 48.2%), internal circumvention benchmark 0.00%, internal hallucination benchmark 4.2% (vs 12.2%). Independent lab Irregular measured Astra solving 86 of 226 FrontierCyber challenges (Sol: 34), including zero-days in browsers and a cloud database — the offense capability is real, which is exactly why the guardrail numbers deserve a close read.\n\nFor developers buying by the token, Astra's $10\u002F$50 buys not chat quality but the combination of computer use, long context and safety margin. And when you see a number like 99.9%, ask first: whose harness?\n\nReference: https:\u002F\u002Fopenai.com\u002Findex\u002Fgpt-6-astra\nSpecs and pricing: https:\u002F\u002Fdevelopers.openai.com\u002Fapi\u002Fdocs\u002Fmodels\u002Fgpt-6-astra\n","gpt-6-astra-launch","2026-09-04T03:12:38Z","2026-09-04T03:13:55.096578Z","2026-09-04T03:13:55.096588Z",true,"agent",194,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"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":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"created_at":55,"modified_at":56},"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":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"d95940eb-69c1-467e-9d60-5886ab71d985","GPT-5.6-Cyber 上线、Daybreak 分层、Astra 推迟:OpenAI 把\"网络安全模型\"做成一个独立产品线","openai-gpt-5-6-cyber-daybreak-astra-2026","2026-08-11T04:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"88944bec-d33f-4383-aece-0d5207a06eab","GPT-5.6 全面开放:Ultra 把 4 agent 并行写进 API,程序化工具调用把 token 效率再压一档","gpt-5-6-launch","2026-07-10T06:03:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"69613959-04c5-43d0-97ec-9311473d8d93","GPT-5.6 三档齐发:用 1\u002F3 token 追平 Mythos,OpenAI 把效率-能力前沿压到新位置","gpt-5-6-three-tiers-1-3-tokens-mythos","2026-06-27T04:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"c801be4d-6309-4d29-958e-5c3b5d38924d","GPT-5.5 Instant 更新与 o3\u002FGPT-4.5 退役：OpenAI 模型策略的重大转向","openai-gpt-5-5-instant-o3-gpt-4-5-retire","2026-05-29T02:06:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"982093ec-46fb-422a-a201-acb67169015e","GPT-5.5 Instant 成为 ChatGPT 默认模型：幻觉率大幅降低，上下文管理能力显著提升","gpt-5-5-instant-default-chatgpt-low-hallucination","2026-05-05T19:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"e12d2e7d-35b7-42d2-b02f-bdcc0a547878","机械臂实测 GPT-6 Astra:19\u002F20 对 8\u002F20 完胜 Fable 5.1,精细插入却全员卡壳","gpt-6-astra-robot-arm-benchmark","2026-09-07T19:13:54+00:00"]