DATA CUT 2026-07-27 116 活跃模型 680 A 类案例

CASE EVIDENCE / A RECORD

ghostwright / Phantom 使用 GLM-5.1 处理软件工程任务执行

ghostwright / Phantom 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 2026-06-27T09:22:50Z。

原始记录:Phantom adds Z.AI GLM-5.1 as a BYOM backend for an AI co-worker that operates its own computer

A

Chinese Brief

中文案例导读

ghostwright / Phantom 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 2026-06-27T09:22:50Z。 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。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。

厂商Z AI / GLM
模型GLM-5.1
任务类型代码代理与软件工程
审核状态auto_approved

任务

真实任务背景

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:Phantom is an open-source AI co-worker that uses an LLM backend to operate a computer, perform coding and automation work, and can be configured to use Z.AI's Anthropic-compatible API with GLM-5.1.

代码代理与软件工程公开代码库A 类可核验real_case
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:The public repository ships a working agent product with BYOM configuration; its README lists Z.AI GLM-5.1 and GLM-4.5-Air as supported provider models and describes switching providers through a YAML block.

模型作用

GLM-5.1 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-5.1 serves as one of Phantom's selectable reasoning/coding backends for tool-using computer-control workflows, positioned by the project as a lower-cost coding-quality alternative to Claude Opus.

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:Evidence is a public product repository and README showing explicit GLM-5.1 support; it does not publish production usage metrics or a named end-customer deployment.