Model reference · open weights
Zhinao-ChineseModernBert-Embedding is an open-weight embedding model from qihoo360. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.
Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.
What it is
| Released by | qihoo360 |
|---|---|
| Type | Embedding models |
| Task | Embeddings |
| Parameters (lead) | 227M |
| Context | 8k tokens |
| Runs with | sentence-transformers |
| Released | 2026-03-18 |
| Popularity | 1k downloads / month |
| Licence | Open weights |
About
Zhinao-ChineseModernBert系列是针对高推理速度要求、严苛内存限制的工业级场景,从头预训练的中文Base级基座模型与语义嵌入模型。本系列基于ModernBert高效架构与Qwen2Tokenizer分词器,依托超大规模中英文语料完成全流程预训练,在保持Base级参数量(除Embedding外约100M参数)轻量化优势的同时,实现了对同量级模型的全面超越,甚至性能优于更大参数量的主流模型,为中文NLP理解任务、语义检索、向量数据库、RAG检索增强等场景提供高性价比的开箱即用解决方案。
本项目包含两个核心模型:
基于1T Tokens高质量中英文语料完成预训练,以中文语料为核心(占比超65%),辅以英文语料,全面通用互联网、科技、金融、医疗、法律、教育、代码等多领域场景,模型语义理解能力与跨域泛化性远超同量级模型。
| 模型全称 | 核心定位 | 预训练数据规模 | 最大序列长度 | 核心适用场景 | 核心优势 |
|---|---|---|---|---|---|
| Zhinao-ChineseModernBert | 通用中文理解基座 | 1T Tokens 中英文MLM预训练 | 1536 | 文本分类、实体识别、情感分析、关系抽取、长文本理解等各类NLU任务 | Base级参数量,长文本友好,推理速度快,内存占用低 |
| Zhinao-ChineseModernBert-Embedding | 专业中文语义嵌入模型 | 1T Tokens RetroMAE预训练 + 检索预训练 + 中英文MTEB微调 | 512 | 语义检索、向量数据库、RAG检索增强、文本相似度计算、聚类、重排序等场景 | CMTEB Base级参数量 SOTA,向量表征精度高,中英夹杂适配性强,检索性能领先 |
SentenceTransformer(
(0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
(1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
(2): Normalize()
)
本模型在Zhinao-ChineseModernBert通用基座的基础上,针对语义表征场景完成两阶段Embedding训练,实现从通用语义理解到精准向量表征的能力跃迁:
Zhinao-ChineseModernBert在CLUE基准榜单上,实现了以Base级参数量(除Embedding外约100M参数)综合性能超越RoBERTa-wwm-large等大参数量模型,为资源有限的业务场景提供更多可选项。
| model | params | afqmc | tnews | iflytek | cmnli | wsc | csl | ocnli | c3 | mean |
|---|---|---|---|---|---|---|---|---|---|---|
| ChineseModernBERT(large) | ~310M | 73.87 | 56.90 | 60.15 | 83.96 | 52.10 | 86.20 | 79.10 | 82.65 | 71.87 |
| RoBERTa-wwm-large | ~310M | 76.55 | 58.61 | 62.98 | 82.12 | 74.60 | 82.13 | 78.20 | 73.82 | 73.63 |
| RoBERTa-wwm-ext(base) | ~90M | 74.04 | 56.94 | 60.31 | 80.51 | 67.80 | 81.00 | 74.72 | 66.50 | 70.23 |
| Zhinao-ChineseModernBERT(base) | ~110M | 76.99 | 57.51 | 59.56 | 83.82 | 78.95 | 85.70 | 79.08 | 75.42 | 74.63 |
Zhinao-ChineseModernBert-Embedding在CMTEB基准榜单上,登顶Base级参数量(除Embedding外约100M参数)模型最优排名,综合性能超越Qwen3-Embedding-0.6B等主流大参数量嵌入模型。
| Model Name | Params | Dimension | Classification | Clustering | Pair Classification | Reranking | Retrieval | STS | Mean Task Type |
|---|---|---|---|---|---|---|---|---|---|
| bge-base-zh-v1.5 | ~90M | 1024 | 71.79 | 47.49 | 73.18 | 65.02 | 69.41 | 51.67 | 63.07 |
| piccolo-base-zh | ~90M | 768 | 70.05 | 47.12 | 70.17 | 66.68 | 71.20 | 54.39 | 63.27 |
| setlla-base-zh-v3-1792d | ~90M | 1792 | 74.40 | 53.29 | 82.50 | 67.84 | 72.28 | 61.92 | 68.71 |
| setlla-large-zh-v3.5-1792d | ~310M | 1792 | 74.66 | 54.31 | 82.92 | 68.45 | 73.52 | 51.93 | 69.30 |
| Qwen3-Embedding-0.6B | ~596M | 1024 | 71.4 | 68.74 | 76.42 | 62.58 | 71.03 | 54.52 | 67.45 |
| 360Zhinao-Embedding-Base | ~110M | 768 | 73.50 | 65.36 | 86.15 | 67.99 | 69.32 | 60.29 | 70.16 |
注:测评代码与详细测评指标可查看eval_mteb.py与CMTEB leaderboard,模型综合性能实现同量级领先,超越600M参数量的Qwen3-Embedding-0.6B模型。
推荐使用Python 3.10+,核心依赖如下:
pip install torch>=2.6.0 transformers>=4.56.2 sentence-transformers>=5.1.2
我们已将模型权重开源至主流模型平台,可通过以下地址获取:
| 模型 | Hugging Face 地址 |
|---|---|
| Zhinao-ChineseModernBert | 🤗 |
| Zhinao-ChineseModernBert-Embedding | 🤗 |
基
From the published model card. Full card on the HuggingFace links in the sidebar.
Using it via the API
Once AxForge deploys zhinao-chinesemodernbert-embedding for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (zhinao-chinesemodernbert-embedding below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/embeddings \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"zhinao-chinesemodernbert-embedding","input":"text to embed"}'
Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.