可进入对比
档案摘要 / Decision snapshot
先判断能不能用于选型。
有真实任务证据
来自模型卡
边界优先
30 秒结论
可进入选型对比,但仍以案例证据为准。
来自公开模型资料库,并已补厂商/官方/模型家族证据;当前站内暂无可核验 A 类案例。
这页有来源、有风险说明、有公开可核验 A 类案例。
数据可信度 / 证据强度
强
A 类案例
13
AA 评分
42.11
官方来源
2 个补充入口
案例来源
13 条可核验
来自公开资料与模型数据库 + Artificial Analysis + 2 个厂商/官方/模型家族证据入口;AA 不是唯一事实来源。
Evidence Distribution
案例证据分布
GLM-4.7 当前关联 13 条 A 类案例;这里按任务、来源和复核时间观察证据结构。
Evidence Snapshot
证据快照覆盖
GLM-4.7 的 13 条 A 类案例中,已有 10 条完成本地证据快照;3 条仍在快照队列。
可用 manifest 复核
等待落盘
需补抓
需人工复核
证据完整度
5 / 5
Feishu Bitable
2 个入口
13 条
已绑定
已标注
能力边界
用标签表达信号,不用标签替代证据。
通用/待核验
暂无数据
待核验
平台/待核验
中
| 厂商 | Z AI / GLM |
|---|---|
| 厂商页 | Z AI / GLM |
| 发布时间 | 2025-12-22 |
| 模型 ID | glm-4-7 |
| 上下文 | 暂无数据 |
| 输出 | 暂无数据 |
| 模态 | 文本 / 多模态待核验 |
| 推理 | 是 / 推理模型或 thinking 模式 |
| 价格 | 官方未披露 / 暂无数据 |
| 可用平台 | Z AI / GLM |
| AA 评分 | 42.11 |
| 基础模型发布时间 | Artificial Analysis 未披露 |
| 资料来源 | 公开资料、厂商信息和案例库 |
| A 类案例 | 13 |
谱系位置
它在厂商路线中的位置
此页将 GLM-4.7 放入 Z AI / GLM 的当前 Atlas 路线中。若同厂前后代资料不足,先保留为路线追踪入口,不脑补谱系关系。
2025-12-22
GLM-4.7 资料状态
来自公开资料与模型数据库 + Artificial Analysis + 2 个厂商/官方/模型家族证据入口;AA 不是唯一事实来源。
公开档案
模型状态
GLM-4.7 当前标记为已有真实案例。
案例库
公开案例复核
已有 13 条可核验 A 类案例,达到完整补齐线
案例时间线
最近进入档案的 A 类案例
用采集时间呈现证据进入 Atlas 的顺序,方便复核来源新鲜度。
2026-06-27T00:30:00Z
Doclos: AI-powered document automation SaaS extracting structured data from German business PDFs using GLM-4.7-Flash
moriarthur (Doclos) / document_processing
2026-06-27T00:30:00Z
SeekMoney-ai: Multi-platform social media business opportunity finder powered by GLM-4.7 thinking model for deep user pain point analysis
liangdabiao (SeekMoney-ai) / data_analysis
2026-06-27T00:30:00Z
GLM-4.7-Flash on AMD Strix Halo: Local deployment achieving near-Claude-Code speed via llama.cpp ROCm with Vulkan acceleration
ndom91 / software_development
2026-06-27T00:30:00Z
glm-acp-agent: TypeScript ACP agent using GLM-4.7 as reasoning core for IDE-integrated coding assistance with file system and terminal tools
stefandevo / software_development
2026-06-26T23:55:00Z
Perplexity Clone: 基于 GLM-4.7-Flash 的 RAG 搜索引擎
JonniTech / search_and_rag
适合场景
- - 作为完整模型索引和厂商路线追踪入口
- - 推理/agentic workflow 候选
不适合
- - 缺少真实案例时不要包装成推荐
- - 缺字段时不要脑补价格、性能或上下文
真实案例
可核验 A 类案例精选
当前展示 13 条代表案例;排序优先看模型卡精选、证据可信度、快照状态和公开仓库/技术任务信号。案例总数仍为 13。
Representative Evidence
代表案例排序
当前第一代表案例是“Doclos: AI-powered document automation SaaS extracting structured data from German business PDFs using GLM-4.7-Flash”。它会优先出现在本页,是因为该案例同时进入模型卡精选、具备 100/100 的证据可信度,并带有 已快照证据。
模型卡精选优先
A+ 完整链路
archive/evidence/glm-4-7-doclos-document-automation/manifest.json
Coding / repo / agent
moriarthur (Doclos) 使用 GLM-4.7 处理知识检索和问答
moriarthur (Doclos) · GLM-4.7
A+ 完整链路 · 代码仓库证据
moriarthur (Doclos) 公开的知识库与检索问答案例,来源为 公开代码库,复核于 2026-06-27T00:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕知识检索和问答的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:Doclos is a SaaS platform that automates business document processing for German Mittelstand (small and medium enterprises). Users upload PDF invoices (Rechnungen), contracts (Verträge), offers (Angebote), and delivery …
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:A production SaaS platform that automatically processes German business documents — invoices, contracts, offers, delivery notes — extracting structured, searchable data from PDFs into a PostgreSQL database with JWT auth…
模型作用:GLM-4.7 在该案例中承担知识检索和问答相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7-Flash serves as the core AI engine in Doclos's extraction pipeline, responsible for understanding document layout and semantics, extracting key fields (dates, amounts, parties, line items) with per-field confide…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Repo is open-source on GitHub with clear README and architecture docs. Model usage is explicitly documented in the tech stack table. No official Zhipu customer story — this is a third-party developer's production build.
原始记录:Doclos: AI-powered document automation SaaS extracting structured data from German business PDFs using GLM-4.7-Flash
stefandevo 使用 GLM-4.7 处理软件工程任务执行
stefandevo · GLM-4.7
A+ 完整链路 · 代码仓库证据
stefandevo 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-27T00:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕软件工程任务执行的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:Developer stefandevo built glm-acp-agent, a TypeScript implementation of the Agent Client Protocol (ACP) that uses Zhipu AI's GLM model family (specifically GLM-4.7 and GLM-5.1) as its reasoning core. The agent connects…
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:An open-source TypeScript agent (23 GitHub stars) that provides ACP-compatible integration of GLM-4.7 into developer IDEs, enabling real-time streaming responses, file system operations, terminal commands, and web acces…
模型作用:GLM-4.7 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7 serves as the primary reasoning engine for the agent's decision-making and tool-calling capabilities. The model's instruction following, function calling support, and long-context handling enable the agent to au…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Open-source GitHub repo with 23 stars. Written in TypeScript, specifically designed for Z.AI GLM Coding Plan. Developer explicitly chose GLM-4.7 for its agentic and tool-coding capabilities. Third-party developer projec…
原始记录:glm-acp-agent: TypeScript ACP agent using GLM-4.7 as reasoning core for IDE-integrated coding assistance with file system and terminal tools
liangdabiao (SeekMoney-ai) 使用 GLM-4.7 处理软件工程任务执行
liangdabiao (SeekMoney-ai) · GLM-4.7
A+ 完整链路 · 代码仓库证据
liangdabiao (SeekMoney-ai) 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-27T00:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕软件工程任务执行的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:SeekMoney-ai is a web application that helps entrepreneurs discover business opportunities from social media by automatically identifying core user pain points. The system collects data from 8 platforms (Douyin, TikTok,…
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:A production web application that processes social media data across 8 platforms, performs AI-driven semantic clustering and deep pain point analysis, outputs structured market opportunity reports with priority scoring …
模型作用:GLM-4.7 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7's thinking model (thinking/reasoning mode) is the core analytical engine that performs deep multi-layer analysis of user pain points extracted from social media conversations. The model goes beyond surface-level…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:Open-source GitHub repo with detailed Chinese README. Uses OpenAI-compatible API pointing to GLM endpoints. This is a third-party developer's product, not an official Zhipu customer story. The project explicitly states it uses 'openai兼容/GLM-4.7 思考模型' (GLM-4.7 thinking model).
原始记录:SeekMoney-ai: Multi-platform social media business opportunity finder powered by GLM-4.7 thinking model for deep user pain point analysis
ndom91 使用 GLM-4.7 处理软件工程任务执行
ndom91 · GLM-4.7
A+ 完整链路 · 代码仓库证据
ndom91 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-27T00:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕软件工程任务执行的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:Developer ndom91 deployed GLM-4.7-Flash locally on a Framework Mainboard with AMD Ryzen AI Max 395+ (Strix Halo) and 128GB RAM, running the model via llama.cpp with ROCm/Vulkan GPU acceleration in a Docker container. Th…
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:A working local deployment of GLM-4.7-Flash on consumer AMD hardware (Framework Mainboard, Ryzen AI Max 395+, 128GB) achieving daily-use performance for coding assistance. The developer reports near-Claude-Code responsi…
模型作用:GLM-4.7 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7-Flash's efficiency and performance at local inference enables a fully offline, private coding assistant experience. The model's quality is high enough that the developer considers it a viable alternative to clou…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Open-source GitHub repo with Docker Compose setup and clear documentation. Developer's subjective speed comparison to Claude Code. Hardware-specific (Strix Halo with ROCm support). This is an individual developer's depl…
原始记录:GLM-4.7-Flash on AMD Strix Halo: Local deployment achieving near-Claude-Code speed via llama.cpp ROCm with Vulkan acceleration
JonniTech 使用 GLM-4.7-Flash 处理知识检索和问答
JonniTech · GLM-4.7-Flash
A+ 完整链路 · 代码仓库证据
JonniTech 公开的知识库与检索问答案例,来源为 公开代码库,复核于 2026-06-26T23:55:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:构建高保真 Perplexity AI 搜索界面克隆,核心 RAG 流水线使用 GLM-4.7-Flash 作为 LLM:用户查询→SerpAPI 获取实时网页搜索结果→上下文提取→GLM-4.7-Flash 生成流式回答并附带引用来源。
公开产物:完整的 RAG 搜索引擎应用,支持实时网页搜索、流式回答、Markdown 渲染、代码高亮、引用来源展示、用户认证(Clerk)和对话历史持久化。React 19 + TypeScript + Tailwind CSS 技术栈。
模型作用:GLM-4.7-Flash 是 RAG 流水线中的核心 LLM,负责接收 SerpAPI 搜索结果构建的上下文,生成准确、带引用的自然语言回答。整个应用的功能正确性直接依赖 GLM-4.7-Flash 的推理和生成能力。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:项目为个人 demo 项目,stars 较少;使用 GLM-4.7-Flash(免费快速版)而非 GLM-4.7 主版本。
原始记录:Perplexity Clone: 基于 GLM-4.7-Flash 的 RAG 搜索引擎
applex250 使用 GLM-4.7 处理文档理解和结构化处理
applex250 · GLM-4.7
A+ 完整链路 · 代码仓库证据
applex250 公开的文档理解与处理案例,来源为 公开代码库,复核于 2026-06-26T23:55:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:使用智谱 GLM-4.7 构建英文学术文献 Markdown 翻译工具,支持 PDF/Word/PPT/图片等多格式输入(通过 MinerU 转换为 Markdown),智能分段、并发翻译、术语一致性保持、表格/公式/代码块过滤与保留,以及 Web 可视化界面。
公开产物:完整的翻译工具,支持命令行和 Web 界面(拖拽上传、任务队列、实时日志),输出高质量中文学术文献 Markdown。过滤特殊元素后翻译可节省 50%+ Token 消耗。v2.1 版本集成 MinerU API 支持多格式直接翻译。
模型作用:GLM-4.7 是翻译流水线的核心 LLM,负责将英文学术文本翻译为中文,同时保持术语一致性和学术语言准确性。智能分段后并发调用 GLM-4.7 API 进行翻译。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:项目为个人开源项目;.env 示例中默认使用 glm-4 而非 glm-4.7,但 README 和项目描述明确标注基于 GLM-4.7。
原始记录:mtrans: 基于 GLM-4.7 的英文学术文献 Markdown 翻译工具
ChuanMeng (SIGIR 2026 研究团队) 使用 GLM-4.7-Flash 处理研究分析和报告生成
ChuanMeng (SIGIR 2026 研究团队) · GLM-4.7-Flash
A+ 完整链路 · 代码仓库证据
ChuanMeng (SIGIR 2026 研究团队) 公开的研究与报告生成案例,来源为 公开代码库,复核于 2026-06-26T23:55:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:SIGIR 2026 论文《Revisiting Text Ranking in Deep Research》使用 GLM-4.7-Flash (30B) 作为深度研究 Agent,在 BrowseComp-Plus 数据集上评估 5 种检索器(BM25、SPLADE-v3、RepLLaMA、Qwen3-Embedding-8B、ColBERTv2)和 3 种重排器(monoT5-3B、RankLLaMA-7B、Rank1-7B)的文本排序性能。
公开产物:发表于 SIGIR 2026(第 49 届 ACM SIGIR 信息检索国际会议)的研究论文,提供了深度研究场景下文本排序方法的全面评估。发布了 BrowseComp-Plus 段落语料库、所有检索器索引和完整执行轨迹数据。
模型作用:GLM-4.7-Flash (30B) 作为两个深度研究 Agent 之一(与 gpt-oss-20b 并列),通过本地 vLLM 服务器部署运行,在多轮检索-重排序循环中执行信息检索和推理任务,是论文实验评估的核心 Agent 模型。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:使用 GLM-4.7-Flash (30B) 而非 GLM-4.7 主版本;属于学术研究实验而非生产应用,但有完整的论文发表和开源代码,证据等级高。
原始记录:SIGIR 2026: GLM-4.7-Flash 作为深度研究 Agent 进行文本排序评估
kittycheeseburger 使用 GLM-4.7-Flash 处理研究分析和报告生成
kittycheeseburger · GLM-4.7-Flash
A+ 完整链路 · 代码仓库证据
kittycheeseburger 公开的研究与报告生成案例,来源为 公开代码库,复核于 2026-06-26T22:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕研究分析和报告生成的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:A front-end/back-end separated emotion analysis chatbot web application that uses GLM-4.7-Flash as the default conversation engine. The system combines GLM-4.7-Flash for dialogue generation with a local Chinese RoBERTa …
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:A full-stack web app (FastAPI backend + Vite/React frontend) that provides real-time emotion-aware chat. GLM-4.7-Flash generates conversational responses, while a local Transformer model classifies user emotions (return…
模型作用:GLM-4.7-Flash 在该案例中承担研究分析和报告生成相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7-Flash is the primary conversation generation model, producing natural dialogue responses in the emotion analysis chatbot. It is called via the Zhipu Open Platform API (open.bigmodel.cn) and configured as the def…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Student/personal project. No deployment URL provided. API key required for GLM backend.
原始记录:Emotion Chatbot: Full-Stack Emotion Analysis Chatbot Powered by GLM-4.7-Flash
graphanov / john-lomein (automated … 使用 GLM-4.7 处理软件工程任务执行
graphanov / john-lomein (automated maintainer) · GLM-4.7
A+ 完整链路 · 代码仓库证据
graphanov / john-lomein (automated maintainer) 公开的代码代理与软件工程案例,来源为 公开代码库,复核于 2026-06-26T22:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕软件工程任务执行的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:LazyGLM is a standalone terminal coding agent CLI for GLM models, providing a Claude Code-style interactive REPL for AI-assisted software development. It routes tasks by type to different GLM model tiers: GLM-5.2 for ha…
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:An npm-installable CLI tool (lazyglm) that provides an interactive agentic coding shell with GLM model routing, session persistence, thinking-token replay across turns, and headless subprocess mode for CI/CD integration…
模型作用:GLM-4.7 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7 is the designated model for routine coding tasks, completion verification, and quick-edit turns within LazyGLM's routing system. It handles the 'verifier' and 'quick' roles, providing efficient, cost-effective c…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:New project (0 stars as of collection). Automated maintainer bot. No formal license stated in README.
原始记录:LazyGLM: GLM-Native Coding Agent CLI with Model Routing to GLM-4.7
tljcpa 使用 GLM-4.7-Flash 处理研究分析和报告生成
tljcpa · GLM-4.7-Flash
A+ 完整链路 · 代码仓库证据
tljcpa 公开的研究与报告生成案例,来源为 公开代码库,复核于 2026-06-26T22:30:00Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:这是一个围绕研究分析和报告生成的真实任务,公开材料可以回溯到具体使用者和具体产物。 原始资料写作:An LLM-powered CLI tool for automated academic paper polishing and Word document formatting. The tool uses GLM-4.7-Flash to perform academic-grade content rewriting, grammar correction, and structural recognition on rou…
公开产物:公开材料提供公开代码、README 或项目配置,可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作:A Python CLI tool (main.py) that takes a rough thesis draft (input.txt), calls GLM-4.7-Flash for academic content polishing via the Zhipu API, applies chunked processing for long documents, and outputs a properly format…
模型作用:GLM-4.7-Flash 在该案例中承担研究分析和报告生成相关的生成、分析、编排或实现角色。 原始资料写作:GLM-4.7-Flash is the default and recommended backend for the paper polishing task, chosen for its domestic accessibility in China, free tier availability (200K context), and strong Chinese text understanding capability.…
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:当前判断基于公开材料;若产物下线、仓库变更或模型参与比例仅来自作者自述,需要在引用前重新复核。 原始资料写作:Student project for undergraduate thesis formatting. Free tier rate-limited (~1 QPS). Requires Zhipu API key with real-name verification.
原始记录:Paper Polisher MVP: Academic Thesis Polishing and Auto-Formatting Tool Using GLM-4.7-Flash
数字游牧人 (Bilibili UP主/程序员) 使用 GLM-4.7 处理软件工程任务执行
数字游牧人 (Bilibili UP主/程序员) · GLM-4.7
A 高可信 · 社区公开记录
数字游牧人 (Bilibili UP主/程序员) 公开的代码代理与软件工程案例,来源为 bilibili,复核于 2026-06-26T17:06:15Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:程序员使用智谱 GLM-4.7 作为 Claude Code 的推理后端,完成实际编程开发任务。视频展示了如何将 GLM-4.7 接入 Claude Code 工具链,利用其编程能力完成代码编写、调试和项目开发。该视频获得 28 万播放和 5404 点赞,证明了大量开发者认可该用法。
公开产物:成功将 GLM-4.7 接入 Claude Code 作为编程后端,完成实际代码开发任务。视频展示了完整的接入配置流程和实际编程效果,证明 GLM-4.7 在 agentic coding 场景下的可用性和效果。
模型作用:GLM-4.7 作为 Claude Code 的推理后端,提供编程理解、代码生成和多步推理能力。其增强的编程能力和长上下文支持使其适合作为 coding agent 的底层模型。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:视频为中文内容,描述基于视频标题和描述的推断;未能提取视频内具体代码输出细节。
原始记录:程序员使用 GLM-4.7 作为 Claude Code 后端进行编程开发
OpenClaw开源社区 使用 GLM-4.7-Flash 处理智能体流程编排
OpenClaw开源社区 · GLM-4.7-Flash
A 高可信 · 社区公开记录
OpenClaw开源社区 公开的智能体工作流案例,来源为 bilibili,复核于 2026-06-26T17:06:15Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:OpenClaw 开源社区通过 1Panel + Ollama 一键部署 GLM-4.7-Flash 本地 AI 智能体,实测 Skills 安装、文件整理、Web 搜索等功能。视频展示了完整的本地部署流程,验证 GLM-4.7-Flash 作为轻量 MoE 模型在本地运行时不占用大量资源,真正实现 Token 自由。视频获得 7.2 万播放和 937 点赞。
公开产物:成功在本地环境部署 GLM-4.7-Flash AI 智能体,实测验证了 Skills 安装、文件整理、Web 搜索等功能的可用性。模型在本地运行轻量不占资源,无需支付 API 费用。
模型作用:GLM-4.7-Flash 作为 30B 级 SOTA MoE 模型,其轻量级架构使其适合本地部署,在保持较高性能的同时显著降低资源消耗。模型的中文理解能力使其在本地智能体场景中表现优异。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:视频为中文内容,具体功能测试细节需参考视频实际内容;OpenClaw 为开源项目而非大型企业。
原始记录:OpenClaw 社区使用 GLM-4.7-Flash 本地部署 AI 智能体实现 Token 自由
御风大世界 (Bilibili UP主/开发者) 使用 GLM-4.7 处理真实任务执行
御风大世界 (Bilibili UP主/开发者) · GLM-4.7
A 高可信 · 社区公开记录
御风大世界 (Bilibili UP主/开发者) 公开的真实任务执行案例,来源为 bilibili,复核于 2026-06-26T17:06:15Z。 Model Atlas 将它标记为 A 类证据,因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。
任务:开发者使用 GLM-4.7 完成原本需要数周工作量的前后端全栈开发任务。视频标题明确表示'GLM-4.7 把我几周的工作量一次性梭哈了',展示了前后端实际测试结果和配套提示词,证明模型在真实全栈项目中的高效生产力。
公开产物:使用 GLM-4.7 一次性完成了原本需要数周的前后端开发工作,视频附带了完整的提示词和前后端实测演示,证明 GLM-4.7 在复杂全栈开发场景下的实际产出能力。
模型作用:GLM-4.7 的增强编程能力和多步推理能力使其能够理解复杂项目需求并生成前后端代码,大幅缩短开发周期。模型在代码理解和生成方面的提升是完成此任务的关键。
A 类理由:有具体使用者、具体任务、公开原始证据和可访问产物,可公开核验。
风险边界:视频为中文内容,具体项目细节和代码质量需参考视频实际内容;未能提取视频内具体代码输出。
原始记录:开发者使用 GLM-4.7 完成数周工作量的前后端全栈开发
数据缺口
只基于现有字段判断缺口;缺失项不会被猜测填充。
- - 补充方向:核验 价格 的官方/API/案例来源。
- - 补充方向:核验 上下文 的官方/API/案例来源。
- - 补充方向:核验 输出 的官方/API/案例来源。
下一步验证建议
- - 继续保留 A 类案例的原始证据、产物页和版本快照。
- - 补官方发布/API/System Card 来源;若官方未披露,继续标注“官方未披露”。
- - 把 benchmark、教程、发布文保留为背景资料,不提升为真实案例。