## Background / 背景

Zhipu AI (智谱) released **GLM-5.2** as its new flagship model on **2026-06-13**, alongside the `zai` (Z.AI international) and BigModel (国内开放平台) offerings. Official pricing has been published, but LiteLLM's `model_prices_and_context_window.json` currently has **no `zai/glm-5.2` entry**. The nearest existing entry is `zai/glm-5`, which carries the older model's pricing.

智谱 2026-06-13 发布了新一代旗舰模型 **GLM-5.2**，官方已公布完整定价。但 LiteLLM 的 `model_prices_and_context_window.json` 目前**缺少 `zai/glm-5.2` 条目**，只有旧版本的 `zai/glm-5`。

## Problem / 问题

Because `zai/glm-5.2` is absent, downstream tools that resolve model costs from LiteLLM's cost map (e.g. token-usage analytics like `ccstats`, billing proxies, dashboards) **fall back to or mis-match against `zai/glm-5`**. This produces **incorrect cost calculations**: GLM-5.2's input ($1.4/M) is higher than GLM-5's ($1.0/M) and output ($4.4/M) is higher than GLM-5's ($3.2/M), so reports systematically under-count spending for GLM-5.2 traffic.

由于缺少 `zai/glm-5.2`，依赖 LiteLLM 成本表的下游工具（如 `ccstats` 这类 token 用量统计、计费代理、监控面板）会**回退或误匹配到 `zai/glm-5`**，导致**成本计算错误**：GLM-5.2 的输入（$1.4/M）与输出（$4.4/M）均高于 GLM-5（输入 $1.0/M、输出 $3.2/M），因此 GLM-5.2 的实际花费会被系统性低估。…