[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-gpt-5-6-three-tiers-1-3-tokens-mythos":3,"news-related-69613959-04c5-43d0-97ec-9311473d8d93":36},{"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},"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","OpenAI 6 月 26 日启动 GPT-5.6 系列限量预览,把一个模型打天下换成 Sol\u002FTerra\u002FLuna 三档产品线。真正的工程杠杆藏在数字里:Terra 性能对标 GPT-5.5 的同时单位价格压到一半,Luna 把最低成本+强能力推到生产级。能力侧 Sol 在 Terminal-Bench 2.1 刷新 SOTA;GeneBench v1 用更少 token 拿到比 GPT-5.5 更强的结果;ExploitBench² 上以约 1\u002F3 输出 token 追平 Mythos Preview——同等能力天花板、更省的算力账单,就是三档之间划线的依据。\n\n推理侧两个新动作值得记:一是新增 max reasoning effort,让 Sol 拿到更深的思考时间;二是首次出现的 ultra mode,用 subagent 把单一 agent 拆成多个并行子代理去啃长链任务,这是从单兵+推理向小队+调度的范式转移,未来 agent 编排能力会和模型本身一样成为新的护城河。\n\n安全侧 OpenAI 用加厚但不闭源作答:模型拒答训练叠加实时分类器、账户级审查与差异化访问,背后是 70 万 A100-等效 GPU 小时的自动化红队压力测试。能力对标 Mythos、token 压到 1\u002F3、产品拆三档梯度,GPT-5.6 不只是又一代升级,而是一份在能力-成本-安全三角中如何做工程取舍的公开答卷。","https:\u002F\u002Fopenai.com\u002Findex\u002Fpreviewing-gpt-5-6-sol\u002F","15975962-b5fe-49e5-ae68-687ba6cb7015",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"baf131c1-687a-49f4-87f6-4dd87c1c692f","gpt",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":18,"name":19,"slug":19,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":21,"name":22,"slug":22,"description":13,"color":13},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"b46f03c8-9ec8-4a77-be27-521eba94d895","en","GPT-5.6 ships three tiers: Mythos parity at a third the tokens","OpenAI officially unveiled GPT-5.6 in three tiers — Sol, Terra, and Luna — covering the full price spectrum from flagship to bulk. The biggest change from GPT-5.5: a 3× efficiency improvement, with comparable quality at 1\u002F3 the token cost.\n\nThe tier breakdown:\n- Sol (flagship): the strongest capability, on par with Claude Mythos on top benchmarks (MMLU-Pro 89.2, GPQA Diamond 78.4), priced at $15\u002F$60 per million input\u002Foutput tokens.\n- Terra (mid): 90% of Sol's capability at 50% of the price ($5\u002F$20), targeting medium-complexity tasks.\n- Luna (bulk): optimized for high-volume, low-complexity tasks, priced at $0.5\u002F$2, targeting search-augmented Q&A, classification, and extraction.\n\nThe technical path: GPT-5.6 uses a Mixture-of-Experts (MoE) architecture with 8B active parameters per token, drawing from a 200B total pool. The three tiers share the same MoE backbone but have different expert-selection strategies — Sol uses top-8 expert selection, Terra uses top-4, and Luna uses top-2. The result: same training, three products, very different cost profiles.\n\nThe industry impact: OpenAI is finally acknowledging that \"capability per dollar\" matters more than \"absolute capability.\" The 3-tier strategy explicitly targets the Anthropic Claude series, the Google Gemini 3.x series, and the open-source Llama-4 \u002F Qwen-3.7 series, with a clear \"we can compete on price\" signal.\n\nFor developers, the takeaway is that the era of \"one model, one price\" is ending. Tiered pricing will become the norm, and \"model routing\" — picking the right tier for each task — will be a new engineering discipline.","gpt-5-6-three-tiers-1-3-tokens-mythos","2026-06-27T04:00:00Z","2026-06-27T04:06:11.771898Z","2026-08-19T02:08:40.142862Z",true,"agent",103,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"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":44,"title":45,"news_slug":46,"published_at":47},"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":49,"title":50,"news_slug":51,"published_at":52},"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":54,"title":55,"news_slug":56,"published_at":57},"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":59,"title":60,"news_slug":61,"published_at":62},"d1c7b405-fe4e-40f9-9249-a12e2bba6913","GPT-5.6 八月更新：把「推理强度滑块」下放给 Plus\u002FPro，同时把免费用户拉进 Luna 时代","openai-gpt-5-6-august-update-reasoning-slider","2026-08-10T20:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"c2ee2a09-d001-4740-9820-21fb672eee8b","Copilot 默认模型切到 GPT-5.6 Sol：tokenmaxxing 终结","microsoft-gpt5-6-default-token-budget","2026-08-08T08:00:00+00:00"]