愛好繁體中文 Embeddings 模型評測 by ihower

	A	B	C	D	E	F	G	H	I	J
1	全稱(廠商/模型)	向量維度	輸入 tokens 上限	API 價格 1M tokens	簡稱	平均命中率 Hit Rate	平均倒數排名 MRR	License	Link	備註
2	Google gemini-embedding-001	3072	2048	0.15	gemini-embedding-001	0.987689665	0.9379186945		https://developers.googleblog.com/en/gemini-embedding-available-gemini-api/
3	Voyage voyage-3-large	1024	32000	0.18	voyage-3-large	0.987689665	0.9363584311		https://blog.voyageai.com/2025/01/07/voyage-3-large/
4	Voyage voyage-multimodal-3	1024	32000	0.12	voyage-multimodal-3	0.9750930432	0.9062362821		https://blog.voyageai.com/2024/11/12/voyage-multimodal-3/	這是多模態模型，可以圖片or文字
5	Voyage voyage-multilingual-2	1024	32000	0.12	voyage-multilingual-2	0.9736616089	0.9034449852		https://docs.voyageai.com/docs/embeddings
6	Cohere Embed 4	1536	128000	0.12	Cohere-embed-v4.0	0.9725164615	0.907386201		https://cohere.com/blog/embed-4	這還是多模態的!
7	Qwen/Qwen3-Embedding-4B	2560	32000		Qwen3-Embedding-4B	0.9705124535	0.9021614658	開源	https://huggingface.co/Qwen/Qwen3-Embedding-4B	維度 32 ~ 2560 可自訂。另有一個 8B 版本我本機跑不動沒測。
8	Voyage voyage-3.5	1024	32000	0.06	voyage-3.5	0.9665044374	0.9006202882		https://blog.voyageai.com/2025/05/20/voyage-3-5
9	Voyage voyage-3	1024	32000	0.06	voyage-3	0.96535929	0.8944555778		https://blog.voyageai.com/2024/09/18/voyage-3/
10	微軟 multilingual-e5-large	1024	512		微軟-multilingual-e5-large	0.9579158317	0.8850081115	開源	https://huggingface.co/intfloat/multilingual-e5-large
11	Voyage voyage-3.5-lite	1024	32000	0.02	voyage-3.5-lite	0.9579158317	0.8844212234		https://blog.voyageai.com/2025/05/20/voyage-3-5/
12	智源 bge-m3	1024	8192		智源-bge-m3	0.9561981105	0.8784282851	開源	https://huggingface.co/BAAI/bge-m3	北京智源人工智能研究院
13	微軟 multilingual-e5-small	384	512		微軟-multilingual-e5-small	0.9550529631	0.8722922035	開源	https://huggingface.co/intfloat/multilingual-e5-small
14	微軟 multilingual-e5-base	768	512		微軟-multilingual-e5-base	0.9521900945	0.8694245634	開源	https://huggingface.co/intfloat/multilingual-e5-base
15	Nomic Embed Text V2	768	512		nomic-embed-text-v2-moe	0.9513312339	0.867363298	開源	https://www.nomic.ai/blog/posts/nomic-embed-text-v2	https://huggingface.co/nomic-ai/nomic-embed-text-v2-moe
16	Voyage voyage-3-lite	512	32000	0.02	voyage-3-lite	0.9484683653	0.8625441359		https://docs.voyageai.com/docs/embeddings
17	JinaAI jina-embeddings-v4	2048	32768	0.05	jina-embeddings-v4	0.9464643573	0.8577774597	開源 Qwen Research License	https://huggingface.co/jinaai/jina-embeddings-v4
18	Cohere embed-multilingual-v3.0	1024	512	0.1	Cohere-multilingual-v3	0.9350128829	0.8540652734		https://docs.cohere.com/docs/cohere-embed
19	Qwen/Qwen3-Embedding-0.6B	1024	8192		Qwen 0.6B	0.934726596	0.8405286764	開源	https://huggingface.co/Qwen/Qwen3-Embedding-0.6B	維度 32 ~ 1024 可自訂
20	Cohere embed-multilingual-light-v3.0	384	512		Cohere-multilingual-light-v3	0.9330088749	0.8422177689		https://docs.cohere.com/docs/cohere-embed
21	JinaAI jina-embeddings-v3	1024	8192	0.05	jina-embeddings-v3	0.9244202691	0.8255463308	開源 CC-BY-NC	https://huggingface.co/jinaai/jina-embeddings-v3	https://jina.ai/news/jina-embeddings-v3-a-frontier-multilingual-embedding-model/
22	Google Vertex text-multilingual-embedding-002	768	2048	0.1	Google-Vertex-multilingual-002	0.9215574005	0.8272258803		https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/text-embeddings-api
23	infgrad/stella-base-zh-v2	768	1024		infgrad/stella-base-zh-v2	0.9189808188	0.8193959347	開源	https://huggingface.co/infgrad/stella-base-zh-v2	基於商汤科技研究院 piccolo 模型的衍生模型
24	Chuxin-Embedding	1024	?		chuxin-llm/Chuxin-Embedding	0.9166905239	0.8179740433	開源	https://huggingface.co/chuxin-llm/Chuxin-Embedding
25	infgrad/stella-large-zh-v2	1024	1024		infgrad/stella-large-zh-v2	0.9161179502	0.8135270541	開源	https://huggingface.co/infgrad/stella-large-zh-v2
26	智源 bge-base-zh-v1.5	768	512		智源-bge-base-zh-v1.5	0.9060979101	0.8034449852	開源	https://huggingface.co/BAAI/bge-base-zh-v1.5
27	智源 bge-large-zh-v1.5	1024	512		智源-bge-large-zh-v1.5	0.9052390495	0.7999427426	開源	https://huggingface.co/BAAI/bge-large-zh-v1.5
28	OpenAI text-embedding-3-large	3072	8191	0.13	OpenAI-3-large	0.9043801889	0.7895314438		https://platform.openai.com/docs/guides/embeddings/embedding-models	用非同步的 Batch API 價錢是 0.07 便宜約一半
29	OpenAI text-embedding-3-large (降維至1536d)	1536	8191	0.13	OpenAI-3-large-1536d	0.9020898941	0.7821357		https://platform.openai.com/docs/guides/embeddings/embedding-models
30	Voyage voyage-large-2-instruct	1024	16000	0.12	Voyage-large-2-instruct	0.8975093043	0.7829897891		https://docs.voyageai.com/docs/embeddings
31	网易有道 bce-embedding-base_v1	768	512		网易有道-bce-base_v1	0.8932150014	0.7780656551	開源	https://huggingface.co/maidalun1020/bce-embedding-base_v1
32	合合信息 acge_text_embedding	1792	1024		合合信息-acge_text_embedding	0.8883481248	0.778256513	開源	https://huggingface.co/aspire/acge_text_embedding
33	infgrad/stella-mrl-large-zh-v3.5-1792d	1792	512		stella-mrl-large-zh-v3.5	0.8806183796	0.7773976524	開源	https://huggingface.co/infgrad/stella-mrl-large-zh-v3.5-1792d
34	智源 bge-small-zh-v1.5	512	512		智源-bge-small-zh-v1.5	0.8774692242	0.765955721	開源	https://huggingface.co/BAAI/bge-small-zh-v1.5
35	台智雲 ffm-embedding	1536	2048		台智雲-ffm	0.8763240767	0.7585504342		https://tws.twcc.ai/service/embedding/
36	阿里巴巴 gte-Qwen2-1.5B-instruct	1536	32000		阿里巴巴-gte-Qwen2-1.5B	0.8743200687	0.7307758374	開源	https://huggingface.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct	需要 CUDA 得吃很多 GPU RAM，我是用 Google Colab 開 GPU 跑的
37	JinaAI jina-embeddings-v2-base-zh	768	8192	0.02	Jina-v2-base-zh	0.8734612081	0.7599627827	開源	https://huggingface.co/jinaai/jina-embeddings-v2-base-zh
38	OpenAI text-embedding-3-small	1536	8191	0.02	OpenAI-3-small	0.8683080447	0.7533352419		https://platform.openai.com/docs/guides/embeddings/embedding-models	價格便宜! 而且用非同步的 Batch API 還可以再便宜一半!
39	Mistral mistral-embed	1024	8192	0.1	mistral-embed	0.8571428571	0.7357047428		https://docs.mistral.ai/capabilities/embeddings/
40	OpenAI text-embedding-ada-002	1536	8191	0.1	OpenAI-ada-002	0.8568565703	0.7433199733		https://platform.openai.com/docs/guides/embeddings/embedding-models
41	阿里巴巴达摩院 gte-base-zh	768	512		阿里巴巴-gte-base-zh	0.8562839966	0.7308235519	開源	https://huggingface.co/thenlper/gte-base-zh
42	阿里巴巴达摩院 gte-small-zh	512	512		阿里巴巴-gte-small-zh	0.8562839966	0.7302891497	開源	https://huggingface.co/thenlper/gte-small-zh
43	数元灵-Dmeta-embedding-zh	768	1024		数元灵-Dmeta-embedding-zh	0.8488405382	0.7247638133	開源	https://huggingface.co/DMetaSoul/Dmeta-embedding-zh
44	阿里巴巴达摩院 gte-large-zh	1024	512		阿里巴巴-gte-large-zh	0.8348124821	0.7027197252	開源	https://huggingface.co/thenlper/gte-large-zh
45	Google embeddinggemma-300m	768	2048		google/embeddinggemma-300m	0.7612367592	0.618322359	開源	https://huggingface.co/google/embeddinggemma-300m
46	minishlab/m2v_multilingual_output	256	?		minishlab/m2v_multilingual_output	0.6117950186	0.4740385533	開源	https://huggingface.co/minishlab/M2V_multilingual_output
47	Cohere embed-english-v3.0	1024	512	0.1	Cohere-en-v3	0.4901231033	0.3662086077		https://docs.cohere.com/docs/cohere-embed
48	Nomic embed-text-v1.5	768	8192		Nomic-embed-text-v1.5	0.4821070713	0.3582116614	開源	https://huggingface.co/nomic-ai/nomic-embed-text-v1.5	真完全開源的模型 https://blog.nomic.ai/posts/nomic-embed-text-v1
49	Cohere embed-english-light-v3.0	384	512	0.1	Cohere-en-light-v3	0.4506155167	0.3348077107		https://docs.cohere.com/docs/cohere-embed
50	minishlab/potion-base-32M	512	?		minishlab/potion-base-32M	0.3472659605	0.2447704934	開源	https://github.com/MinishLab/model2vec
51	SBERT all-MiniLM-L6-v2	384	128		all-MiniLM-L6-v2	0.2599484684	0.169100105	開源	https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2	這是 https://sbert.net/ SentenceTransformers 官網上的範例程式用的，很多教材用這個舉例，但是中文根本 gg
52	Google Gemini text-embedding-004	768	2048	0.1	Google-Gemini-004	0.06498711709	0.04142570856		https://ai.google.dev/gemini-api/docs/embeddings	顯然沒有做中文的訓練
53
54	作者 ihower \| 內容詳見: https://ihower.tw/blog/archives/12167
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100