https://waytoclawearn.com/zh/news/llama-cpp-decision-models-systemone-five-open-gguf-2026-10 WayToClawEarn zh 2026-10-03T20:17:44.455766Z llama.cpp 原生支持决策模型：5 款开源 GGUF 上线，Agent 路由怎么降本？ https://waytoclawearn.com/zh/news/aleph-alpha-kolibri-78b-open-weight-sovereign-moe-2026-10 WayToClawEarn zh 2026-10-03T20:17:40.390956Z Aleph Alpha 开源 Kolibri：78B 参数 3.46B 激活的德英 MoE，主权 AI 怎么选？ https://waytoclawearn.com/zh/news/xiaomi-mimo-v2-6-pro-agent-arena-top-5-open-source-confirmed-success-2026-10 WayToClawEarn zh 2026-10-03T03:56:24.556694Z 小米 MiMo-V2.6-Pro 登 Agent Arena 开源第五：确认成功率开源第二，国产 Agent 模型怎么选？ https://waytoclawearn.com/zh/news/langchain-model-router-open-swe-64-percent-cost-cut-2026-10 WayToClawEarn zh 2026-10-03T03:56:18.563045Z LangChain 开源模型路由器：Agent 编程中位成本降 64%，质量不降，怎么搭？ https://waytoclawearn.com/zh/news/karpathy-llm-output-formats-ste100-diagrams-html-video-2026-10 WayToClawEarn zh 2026-10-02T23:17:07.602935Z Karpathy 提出 LLM 输出四层阶梯：STE100、图、HTML、视频，Agent 自主工作后人类怎么读懂？ https://waytoclawearn.com/zh/news/overmind-open-source-slm-fine-tuning-platform-2026-10 WayToClawEarn zh 2026-10-02T23:17:03.418203Z Overmind 开源专用小模型训练平台：前情报官创业，用自有数据微调 SLM 怎么打大模型？ https://waytoclawearn.com/zh/news/comfy-agent-autonomous-workflow-builder-comfyui-2026-10 WayToClawEarn zh 2026-10-02T23:16:59.546551Z Comfy Org 发布 Comfy Agent：在画布上自主构建 ComfyUI 工作流，AI 图像流水线怎么搭？ https://waytoclawearn.com/zh/news/ant-group-ling-3-1-flash-560b-moe-open-source-agent-2026-10 WayToClawEarn zh 2026-10-02T19:09:32.058173Z 蚂蚁百灵 Ling-3.1-flash 发布：560B 参数 25B 激活、百万上下文、试用后开源，国产 Agent 模型怎么选？ https://waytoclawearn.com/zh/news/black-forest-labs-flux-3-image-multi-step-editing-bounding-box-2026-10 WayToClawEarn zh 2026-10-02T15:02:47.264579Z Black Forest Labs 发布 Flux 3 Image：单端点做多步局部编辑，AI 图像生成怎么选？ https://waytoclawearn.com/zh/news/aws-strands-decider-2b-open-source-decision-model-agent-routing-2026-10 WayToClawEarn zh 2026-10-02T15:01:07.353437Z AWS 开源 Strands Decider 2B 决策模型：不生成文本只返回概率，AI Agent 路由怎么降本？