Gemini Embedding 001 is Google's generally available text embedding model, with a strong position on the Massive Text Embedding Benchmark (MTEB) Multilingual leaderboard across retrieval, classification, and other tasks, with Matryoshka-based dimension flexibility and support for over 100 languages.
import { embed } from 'ai';
const result = await embed({ model: 'google/gemini-embedding-001', value: 'Sunny day at the beach',})Gemini Embedding 001 reached general availability on May 20, 2025 after ranking highly on the Massive Text Embedding Benchmark (MTEB) Multilingual leaderboard since its experimental launch. Gemini Embedding 001 shows stronger retrieval and classification results than Google's earlier text embedding models on MTEB Multilingual across tasks spanning domains including science, legal, finance, and coding.
The model supports over 100 languages and accepts up to 2,048 input tokens per request. Its MTEB Multilingual leaderboard position reflects evaluations across diverse languages and task types, meaning the benchmark advantage spans cross-lingual retrieval and multilingual classification, not just English-language retrieval.
A key architectural feature is Matryoshka Representation Learning (MRL), which lets you reduce output vector dimensions from the default 3,072 to smaller sizes. Google recommends 3,072, 1,536, or 768 dimensions for the highest quality. This flexibility helps balance retrieval accuracy against vector database storage costs. A corpus embedded at 768 dimensions uses significantly less storage than the same corpus at 3,072, with a measurable but bounded quality tradeoff.
The model uses the embed_content endpoint and costs $0.15 per million input tokens. A free tier is available for experimentation, with higher rate limits in the paid tier.