Skip to content

i18n(zh): token translated as 令牌 reads oddly in LLM context #40976

Closed

Description

@Speechlessmanbilibili

Description

In the Simplified Chinese (zh) locale, the context usage panel translates "token" as 「令牌」 (literally "command tablet"). That term comes from the API credential context (访问令牌 / refresh token), and in an LLM context it reads oddly — e.g. 「推理令牌」 (reasoning token) and 「缓存令牌」 (cache token).

For reference:

  • The Traditional Chinese locale (zht.ts) keeps the original word: 总 token, 输入 token, 推理 token
  • Earlier versions of the zh locale also kept "token"; it was changed to 「令牌」 in a recent translation pass
  • .opencode/glossary/zh-cn.md has no entry for "token"

Suggested fix: use 「词元」 (the common scholarly translation for LLM tokens), or keep the English "token" like the zht locale does.

Plugins

None

OpenCode version

Desktop, latest dev build (2026-08)

Steps to reproduce

  1. Open OpenCode Desktop, set language to 简体中文
  2. Open any session and look at the context usage panel
  3. See 令牌总数 / 输入令牌 / 输出令牌 / 推理令牌 / 缓存令牌 / 令牌数

Screenshot and/or share link

N/A

Operating System

Windows 11

Terminal

N/A (Desktop app)

Activity

  1. tp20c5 commented on Aug 11, 2026

    @tp20c5

    词元确实比令牌更恰当,毕竟token作为令牌翻译时是用在几乎完全不同的场合。但念了好几年的token突然因为一纸正确就要换成词元,一旦注意到了还真有点甩不开的违和感,就跟走在大街上突然有人喝住要你抬腿时先迈左脚一样。

    接受不了的就换繁中🌐️吧,那儿还用token——无关什么正确,因为翻到日文界面是用片假名トークン 时老子也绷不住了😄️,从这一点看能有合适的中文翻译其实也未尝不是好事。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions