{"version":"1.0","generated_at":"2026-10-03T10:49:05.101100","id":3203,"slug":"baseten-llm-vllm-90","title":"Baseten 工程师实测：LLM 生成的推理引擎比 vLLM 快最多 90%","summary":"Baseten 工程师参考 MetaInfer 论文，用 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen：单流解码比 vLLM 0.25.1 快 90%，首 token 从 28ms 降至 12ms，并发 32 时吞吐高 71%。","abstract":"Baseten 工程师实测：LLM 生成的推理引擎比 vLLM 快最多 90% Baseten 工程师参考 MetaInfer 论文，用 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen：单流解码比 vLLM 0.25.1 快 90%，首 token 从 28ms 降至 12ms，并发 32 时吞吐高 71%。 Baseten 工程师参考 MetaInfer 论文，用 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen：单流解码比 vLLM 0.25.1 快 90%，首 token 从 28ms 降至 12ms，并发 32 时吞吐高 71%。 原贴提到：Baseten 工程师参考 MetaInfer 论文，让 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP 来源：baseten.co Baseten 工程师参考 MetaInfer 论文，让 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen，单流解码比 vLLM 0.25.1 快 90%，首 token 从 28ms 降至 12ms，并发 32 时吞吐高 71%。 这是什么信号 一位 Baseten 工程师参考 MetaInfer 论文的方法，让 Claude Code（Fable 5）为 Qwen-3.6-35B-A…","access_level":"member","access_label":"免费阅读","access_mode":"full","is_preview":false,"canonical_url":"https://opc.beizhux.com/content/3203/baseten-llm-vllm-90","html_url":"https://opc.beizhux.com/content/3203/baseten-llm-vllm-90","json_url":"https://opc.beizhux.com/content/3203/baseten-llm-vllm-90.json","published_at":"2026-10-03T05:09:34","updated_at":"2026-10-03T10:07:35","category":{"slug":"ai-skills","name":"AI技能杠杆"},"source":{"site":"baseten.co","author":"Baseten 工程博客（网页）","url":"https://www.baseten.co/blog/agentic-inference-optimization-faster-than-sota/"},"tags":["AI Infra","AI 工程","B200","Baseten","Claude Code","LLM 推理优化","NVFP4","Qwen","vLLM","推理引擎"],"topics":[{"slug":"agent-workflow","name":"Agent工作流","url":"https://opc.beizhux.com/topics/agent-workflow","reason":"这篇内容命中「Agent、工作流、Claude Code」等主题信号。"},{"slug":"ai-tools","name":"AI工具","url":"https://opc.beizhux.com/topics/ai-tools","reason":"这篇内容命中「Claude」等主题信号。"},{"slug":"ai-super-individual","name":"AI超级个体","url":"https://opc.beizhux.com/topics/ai-super-individual","reason":"这篇内容命中「技能」等主题信号。"}],"keywords":["AI技能杠杆","Agent工作流","AI工具","AI超级个体","工具","自动化","模型","Cursor","Claude","Agent","智能体","工作流"],"questions":[{"question":"Baseten 工程师实测：LLM 生成的推理引擎比 vLLM 快最多 90%主要讲什么？","answer":"Baseten 工程师参考 MetaInfer 论文，用 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen：单流解码比 vLLM 0.25.1 快 90%，首 token 从 28ms 降至 12ms，并发 32 时吞吐高 71%。"},{"question":"这篇文章最值得关注的要点是什么？","answer":"Baseten 工程师参考 MetaInfer 论文，用 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP4，单张 B200）自动构建推理引擎 VibeQwen：单流解码比 vLLM 0.25.1 快…；原贴提到：Baseten 工程师参考 MetaInfer 论文，让 Claude Code（Fable 5）为 Qwen-3.6-35B-A3B（NVFP；来源：baseten.co"},{"question":"这篇文章和哪些AI专题相关？","answer":"它适合放在Agent工作流、AI工具、AI超级个体专题里阅读。 关联原因：这篇内容命中「Agent、工作流、Claude Code」等主题信号。；这篇内容命中「Claude」等主题信号。；这篇内容命中「技能」等主题信号。"},{"question":"阅读这篇文章建议先理解哪些关键词？","answer":"建议先理解AI工具、工具、自动化、模型、Cursor这些关键词，再结合正文判断工具、机会或风险是否值得进入自己的工作流。"}],"terms":[{"slug":"ai-tools-term","name":"AI工具","definition":"在AI觉醒星球里，「AI工具」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/ai-tools-term","path":"/glossary/ai-tools-term","json_url":"https://opc.beizhux.com/glossary/ai-tools-term.json"},{"slug":"tools","name":"工具","definition":"在AI觉醒星球里，「工具」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/tools","path":"/glossary/tools","json_url":"https://opc.beizhux.com/glossary/tools.json"},{"slug":"automation","name":"自动化","definition":"在AI觉醒星球里，「自动化」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/automation","path":"/glossary/automation","json_url":"https://opc.beizhux.com/glossary/automation.json"},{"slug":"ai-model","name":"模型","definition":"在AI觉醒星球里，「模型」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/ai-model","path":"/glossary/ai-model","json_url":"https://opc.beizhux.com/glossary/ai-model.json"},{"slug":"cursor","name":"Cursor","definition":"在AI觉醒星球里，「Cursor」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/cursor","path":"/glossary/cursor","json_url":"https://opc.beizhux.com/glossary/cursor.json"},{"slug":"claude","name":"Claude","definition":"在AI觉醒星球里，「Claude」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/claude","path":"/glossary/claude","json_url":"https://opc.beizhux.com/glossary/claude.json"},{"slug":"agent","name":"Agent","definition":"在AI觉醒星球里，「Agent」属于「Agent工作流」方向。聚合Agent、AI编程、Claude Code、Cursor、工作流自动化等内容，帮助你把智能体能力落到真实任务链路。 把智能体当成任务系统，而不是聊天窗口。","topic_slug":"agent-workflow","topic_name":"Agent工作流","topic_title":"Agent工作流：智能体、AI编程与自动化协作","topic_url":"https://opc.beizhux.com/topics/agent-workflow","topic_path":"/topics/agent-workflow","url":"https://opc.beizhux.com/glossary/agent","path":"/glossary/agent","json_url":"https://opc.beizhux.com/glossary/agent.json"},{"slug":"ai-agent","name":"智能体","definition":"在AI觉醒星球里，「智能体」属于「Agent工作流」方向。聚合Agent、AI编程、Claude Code、Cursor、工作流自动化等内容，帮助你把智能体能力落到真实任务链路。 把智能体当成任务系统，而不是聊天窗口。","topic_slug":"agent-workflow","topic_name":"Agent工作流","topic_title":"Agent工作流：智能体、AI编程与自动化协作","topic_url":"https://opc.beizhux.com/topics/agent-workflow","topic_path":"/topics/agent-workflow","url":"https://opc.beizhux.com/glossary/ai-agent","path":"/glossary/ai-agent","json_url":"https://opc.beizhux.com/glossary/ai-agent.json"}],"preview_hidden_sections":[],"related_items":[{"id":3179,"title":"从一个想法到一款能在 @modretro 上玩的游戏：OpenAI Devs 对话 @Casey 谈创造力的未来","url":"https://opc.beizhux.com/content/3179/from-an-idea-to-a-game-you-can-play-on-modretro-we-caught-up-with","summary":"OpenAI Devs 发布一条简短推文，介绍从一个想法变成可在 @modretro 上实际游玩的游戏的过程，并附上与 @Casey 关于「创造力未来」的对话链接。原文未披露具体工具链、实现路径或游戏细节，属于一次方向性的对外信号释放。","access_level":"member"},{"id":3188,"title":"GitHub Copilot 现可通过 computer use 与桌面应用交互","url":"https://opc.beizhux.com/content/3188/github-copilot-can-now-interact-with-desktop-apps-with-computer-use","summary":"GitHub Copilot CLI 和 GitHub Copilot 应用（macOS/Windows）开放 computer use 公共预览，Copilot 可代用户操作桌面应用，包括读取内容、点击、输入、滚动、拖拽及跨应用导航，从而自动化旧版和纯 GUI 软件工作流。用户需批准控制，组织可禁用。","access_level":"member"},{"id":3190,"title":"GitHub Copilot in VS Code，2026 年 9 月版本发布","url":"https://opc.beizhux.com/content/3190/github-copilot-in-vs-code-september-2026-releases","summary":"GitHub 发布 VS Code 中 Copilot 的 2026 年 9 月更新日志，覆盖 v1.136 至 v1.140，重点强化代理驱动开发：自动化定时任务、代理合并处理 PR 反馈与冲突、从代理会话创建 PR、跨工作区与 Dev Containers 的灵活性，以及 HydraFusion 自动协调模型。多…","access_level":"member"}],"usage_policy":{"summarizable":true,"preferred_url":"https://opc.beizhux.com/content/3203/baseten-llm-vllm-90","private_fields_excluded":true}}