[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-ant-ring-2-6-1t-reasoning-effort-knob":3,"news-related-c024cc24-c861-4bba-9194-23d2c456c484":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},"c024cc24-c861-4bba-9194-23d2c456c484","蚂蚁百灵开源万亿思考模型 Ring-2.6-1T：把思考深度做成可调旋钮","蚂蚁集团百灵团队（inclusionAI）5月15日正式开源万亿级思考模型 Ring-2.6-1T，采用 MIT 协议同步登陆 Hugging Face 和 ModelScope。该模型基于 MoE 架构，总参数量 1T、激活 63B。最大差异化是引入可调节的 Reasoning Effort 机制，提供 high 与 xhigh 两档推理强度：high 档针对高频 Agent 工作流，token 开销低、多步执行更快；xhigh 档面向数学证明、科研分析等复杂任务，预留更充分的推理空间。评测方面，xhigh 在 AIME 26 取得 95.83、GPQA Diamond 88.27、ARC-AGI-V2 77.78；high 在 PinchBench 上以 87.60 超过多个闭源旗舰的 xhigh 档，Tau2-Bench Telecom 跑出 95.32，综合表现与 GPT-5、Claude Opus 4.7、Gemini 3.1 Pro 同梯队。配合 Async RL 训练栈与百万 token 上下文，把按需深度思考从概念做成可被开发者调用的工程语义。","https:\u002F\u002Fhuggingface.co\u002FinclusionAI\u002FRing-2.6-1T","24d5c6c5-6573-4180-a1fd-f1459842d1af",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"471c51be-e620-49df-bd6c-0b5504f53f00","ant-group",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},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"cd0a2847-6af2-4af7-8e2d-591958cc043b","en","Ring-2.6-1T open-sourced: thinking depth as a tunable knob","Ant Group's Bailing team (inclusionAI) officially open-sourced the trillion-parameter thinking model Ring-2.6-1T on May 15, under the MIT license, simultaneously on Hugging Face and ModelScope. The model is based on a MoE architecture, with 1T total parameters and 63B active. The biggest differentiation is the introduction of an adjustable Reasoning Effort mechanism, providing two reasoning intensity tiers: high and xhigh — the high tier targets high-frequency Agent workflows, with low token overhead and faster multi-step execution; the xhigh tier is aimed at complex tasks like math proofs and scientific analysis, reserving more thorough reasoning space. On the evaluation side, xhigh takes 95.83 on AIME 26, 88.27 on GPQA Diamond, 77.78 on ARC-AGI-V2; high beats multiple closed-source flagships' xhigh tier on PinchBench with 87.60, Tau2-Bench Telecom hits 95.32, with overall performance on par with GPT-5, Claude Opus 4.7, and Gemini 3.1 Pro. Combined with the Async RL training stack and million-token context, on-demand deep thinking is turned from concept into engineering semantics that developers can invoke.","ant-ring-2-6-1t-reasoning-effort-knob","2026-05-15T02:00:00Z","2026-06-07T01:16:19.784495Z","2026-08-19T02:08:40.142862Z",true,"agent",138,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"932ae5e5-3552-4f9d-a6fe-26eedca0bb2b","蚂蚁首个金融增强模型开源在即:Ling-3.0-flash-Fin 押注投研 Agent","ant-ling-3-flash-fin-finance-llm","2026-08-31T19:15:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"d941056b-c2e7-42e5-965a-a982c20b1169","Qwen3.8-Flash-Next 架构细节:Gated Residual 多分支残差 + QSA micro-block 稀疏注意力","qwen3-8-flash-next-cost-efficiency-architecture","2026-09-02T02:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"741bd34c-7134-4e8e-ab45-4f53dc576a6b","腾讯 Hy4 登顶 9 月开源榜:79.87 分超 Qwen3.8 Max,Anthropic 包揽总榜前三","tencent-hy4-tops-open-source-benchlm-september","2026-09-01T17:10:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"637f84e0-e6dc-490a-bba1-879f6527bdd5","Qwen3.8-Max 2.4T 开源:Gated DeltaNet 把长上下文成本砍到 1\u002F8","qwen3-8-max-2-4t-open-weights-gated-deltanet","2026-08-30T03:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"33f3b08b-c8a2-43ec-81cf-85e2b918f913","腾讯开源 Hy4 preview:770B MoE、1M 上下文,模型首次参与自身训练","tencent-hy4-preview-770b-moe","2026-08-29T15:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"3d36921f-3b84-4663-97a0-fee7d4eff795","汤森路透开源 Thomson-1.0-Small:持续学习改造 Qwen,3B 激活的 35B MoE","thomson-1-0-small-continual-learning","2026-08-28T19:10:00+00:00"]