IT之家 9 月 30 日消息，哔哩哔哩 Index LLM 团队今日正式发布 Index-Translate 模型，2B / 9B / 35B-A3B（preview）文本模型权重在 Hugging Face 与 ModelScope 开放。
Index-Translate 是基于 Qwen3.5 构建的多语言翻译模型家族。文本模型覆盖涵盖中文、英文在内的 150 种语言，兼容术语、格式、保留内容等翻译指令，并将共同的多语基础扩展到语音、音节可控翻译和长文档翻译。
以下为官方精选案例：
原文：狒瘾犯了就去打。在这个游戏社区语境中，“狒瘾”指想玩《最终幻想 XIV》的劲头又上来了。
| 模型 | 英文译文 |
|---|---|
| Index-Translate-9B | When the FFXIV itch hits, just go play. |
| Index-Translate-2B | Go play when my FFXIV addiction kicks in. |
| Hy-MT2-7B | If you get monkey addiction, go fight. |
原文：生活两天，是一种什么感受。同一句话设置三个目标长度，Index-Homura-9B 给出不同措辞：
| 目标 / 实际音节数 | 英文译文 |
|---|---|
| 10 / 10 | to live for two days. What would that be like? |
| 14 / 14 | What would it be like to live there for two days, I wonder? |
| 18 / 18 | What would it be like to live there for two days, trying to get by somehow? |
在一篇约 32K tokens 的奇幻文本中，“王妃”是人物姓名。下表比较原生全文翻译，与同一 9B 模型使用相邻上下文及自动词表的分块翻译。
| 原文位置 | Index-NativeLong-9B 原生全文 | 同一 9B 分块翻译 |
|---|---|---|
| 31.3% | Mentor Wang Fei | Instructor Wangfei |
| 56.6% | Wang Fei | the Dean |
| 84.8% | Wang Fei | The Queen Consort |
IT之家附相关链接：
本文来源：https://www.ithome.com/1/008/914.htm
若有版权问题，请来信告知：13951431913#139.com 或在右方提交评论。
Sep 30, IT Home news. Today, Bilibili's Index LLM team officially released the Index-Translate model. Text model weights for 2B / 9B / 35B-A3B (preview) are open on Hugging Face and ModelScope.
Index-Translate is a multilingual translation model family built on Qwen3.5. The text model covers 150 languages including Chinese and English, is compatible with translation instructions such as terminology, format, and retained content, and extends its shared multilingual foundation to speech, syllable-controllable translation, and long-document translation.
The following are official curated examples:
Original text: “Baboon addiction is acting up, just go and beat it.” In the context of this gaming community, “baboon addiction” refers to the urge to play *Final Fantasy XIV* resurging.
Original text: “What is the feeling of living for two days.” Three target lengths are set for the same sentence; Index-Homura-9B produces different phrasings:
In a fantasy text of approximately 32K tokens, “Princess Consort” is a character name. The table below compares native full-text translation with chunked translation using adjacent context and an automatic vocabulary with the same 9B model.
IT Home attached related links: