Compare embedding models and API pricing
Turn text into vectors for search.
Find embedding models for semantic search, retrieval-augmented generation, recommendations, clustering, and document processing through a consistent API.
33 embedding models are listed today. All of them use one OneInfer API key and pay-as-you-go wallet.
Choose and call a embedding model in 4 steps
The list ranks them by their input price per million tokens. Different billing units are never mixed, so the order is fair.
Lowest comparable price: Google: EmbeddingGemma 300M
Currently the lowest-priced option in the average of available input and output prices per one million tokens group. Different billing units are never mixed.
$0.0022 avg / 1M tokens
View pricing and providersAvailable embedding models
Comparable models are ordered by their current normalized price. Pricing content refreshed November 9, 2026.
33 models
| Model |
|---|
Google: EmbeddingGemma 300MLowest pricedeepinfra EmbeddingGemma 300M is Google DeepMind's small multilingual embedding model for search, classification and clustering. It makes 768-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 2048 tokens.
|
deepinfra BGE Base EN v1.5 is BAAI's base-size English embedding model for retrieval and similarity. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra E5 Base v2 is a base-size English text embedding model trained with weakly supervised contrastive pre-training. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra GTE Base is Alibaba DAMO Academy's base-size English text embedding model for retrieval and similarity. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra all-mpnet-base-v2 is a general-purpose English sentence embedding model for search, clustering and similarity. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra multi-qa-mpnet-base-dot-v1 is an English embedding model trained on 215 million question-answer pairs for semantic search; compare vectors by dot product. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra all-MiniLM-L12-v2 is a small, fast English sentence embedding model. It makes 384-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra all-MiniLM-L6-v2 is a very small, fast English sentence embedding model. It makes 384-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra paraphrase-MiniLM-L6-v2 is a small English sentence embedding model trained to find paraphrases. It makes 384-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra CLIP ViT-B/32 embeds text into CLIP's shared text and image space, for image search by text. It makes 512-dimension vectors, from inputs of up to 77 tokens.
|
deepinfra CLIP ViT-B/32 Multilingual embeds text in 50+ languages into CLIP's shared text and image space. It makes 512-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra text2vec Base Chinese is a Chinese sentence embedding model for similarity and search. It makes 768-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra Qwen3 Embedding 8B is the largest of Alibaba's Qwen3 embedding models, multilingual and strong at retrieval and code search. It makes 4096-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 32768 tokens.
|
deepinfra Qwen3 Embedding 0.6B is the smallest Qwen3 embedding model from Alibaba, for fast multilingual retrieval. It makes 1024-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 32768 tokens.
|
deepinfra Nemotron 3 Embed 1B NVFP4 is the 4-bit build of NVIDIA's compact multilingual text embedding model. It makes 2048-dimension vectors, from inputs of up to 32768 tokens.
|
deepinfra Llama Nemotron Embed VL 1B v2 is NVIDIA's embedding model for retrieval over documents with charts and tables. It makes 2048-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 10240 tokens.
|
deepinfra BGE EN ICL is BAAI's LLM-based English embedding model that improves with few-shot examples in the input. It makes 4096-dimension vectors, from inputs of up to 8192 tokens.
|
deepinfra BGE M3 is BAAI's multilingual embedding model (100+ languages) for long documents. It makes 1024-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 8192 tokens.
|
deepinfra BGE M3 Multi is BAAI's multilingual BGE M3 (100+ languages) for short passages. It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra BGE M3 Multi 8K is BAAI's multilingual BGE M3 (100+ languages) for long documents. It makes 1024-dimension vectors, from inputs of up to 8192 tokens.
|
deepinfra BGE Large EN v1.5 is BAAI's large English embedding model for retrieval and similarity. It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra E5 Large v2 is an English text embedding model trained with weakly supervised contrastive pre-training. It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra Multilingual E5 Large is a text embedding model for 100 languages. Prefix inputs with "query: " or "passage: ". It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra Multilingual E5 Large Instruct is a text embedding model for 100 languages that takes a task instruction with each query. It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra GTE Large is Alibaba DAMO Academy's large English text embedding model for retrieval and similarity. It makes 1024-dimension vectors, from inputs of up to 512 tokens.
|
deepinfra Nemotron 3.5 Embed 1B is NVIDIA's compact embedding model for document retrieval and semantic search. It makes 2048-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 32768 tokens.
|
deepinfra Nemotron 3 Embed 1B is NVIDIA's compact multilingual text embedding model (34 languages) for retrieval. It makes 2048-dimension vectors, from inputs of up to 32768 tokens.
|
deepinfra Qwen3 Embedding 4B is a mid-sized Qwen3 embedding model from Alibaba for multilingual retrieval and code search. It makes 2560-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 32768 tokens.
|
deepinfra EmbeddingGemma 2 is Google DeepMind's 740M embedding model for search, retrieval and clustering. It makes 768-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 8192 tokens.
|
deepinfra Nemotron 3.5 Embed 8B is NVIDIA's embedding model for document retrieval and semantic search. It makes 4096-dimension vectors, or shorter ones on request (dimensions), from inputs of up to 32768 tokens.
|
deepinfra Nemotron 3 Embed 8B is NVIDIA's multilingual text embedding model (34 languages) for retrieval and RAG. It makes 4096-dimension vectors, from inputs of up to 32768 tokens.
|
b0d92f943e1142f4a6fe212f46a15c5e Codestral Embed is Mistral AI's embedding model for code search and retrieval over code repositories. It makes 1536-dimension vectors, or up to 3072 on request (dimensions), from inputs of up to 8192 tokens.
|
ec5d66c0fd0b41aa8f7a06718603c6b8 Mistral Embed is Mistral AI's general-purpose text embedding model for retrieval, classification and similarity search. It makes 1024-dimension vectors from inputs of up to 8192 tokens.
|
How to choose
- RAG
- Semantic search
- Recommendations
- Document clustering
Pricing methodology
OneInfer compares only positive prices with the same billing unit. Per-minute, per-character, per-token, per-image, per-video, and per-second rates remain separate. Prices can change, so the current model page and console remain the source of truth.
Frequently asked questions
How are embedding model prices compared?
Embedding models are compared using their positive input-token price per one million tokens when that pricing is available.
Can I use these models with my existing vector database?
Yes. Store the returned vectors in your preferred vector database and use them for retrieval or similarity search.