Google Gemini Integration Text model, Gémini-Embedding-001is now generally available for developers via the Gemini and Google AI Studio API, providing powerful multilingual and flexible text representation to the wider IA ecosystem.
Multilingual support and dimensional flexibility
- Supports more than 100 languages: Gemini integration is optimized for global applications and operates in more than a hundred languages, making it an ideal solution for projects with various linguistic requirements.
- Learning the representation of Matryoshka: Architecture exploits the learning of the representation of Matryoshka, allowing developers of Integration vector scale Effectively – Choose Dimensions 3072 by default or scales at 1536 or 768, depending on the compromise of your request between precision and performance. This adaptable structure allows you to optimize for speed, cost and storage with minimum quality loss When you reduce the size of the vector.
Technical specifications and model performance
- Entry capacity: Process until 2048 tokens per entrancewith suggestions that future updates could further extend this limit.
- Reference leader: Since its early deployment, Gemini-Embedding-001 has reached best scores On the massive classification of text integration (MTEB), a multilingual classification, exceeding previous Google models and external offers through areas such as science, legal and coding.
- Unified architecture: Consolidates the capacities which previously required several specialized models, simplifying workflows for research, recovery, grouping and classification tasks.
Key characteristics
- Default integration with 3072 dimensions (sustained truncation for 1536 or 768)
- Vector standardization For compatibility with the similarity of the Cosinus and vector research frames
- Minimum performance with reduced dimensionality
- Improved compatibility with popular vector databases (for example, Pinecone, Chromadb, Qdrant, Weavate) and Google Databases (Alloydb, Cloud SQL)
| Metric / task | Gémini-Embedding-001 | Inherited Google models | COHERE V3.0 | OPENAI-3-GARM |
|---|---|---|---|---|
| MTEB (Multilingual) Middle (task) | 68.37 | 62.13 | 61.12 | 58.93 |
| MTEB (Multilingual) Middle (Tasktype) | 59.59 | 54.32 | 53.23 | 51.41 |
| Durex mining | 79.28 | 70.73 | 70.50 | 62.17 |
| Classification | 71.82 | 64.64 | 62.95 | 60.27 |
| Grouping | 54.59 | 48.47 | 46.89 | 46.89 |
| Instant recovery | 5.18 | 4.08 | -1.89 | -2.68 |
| Multilabel classification | 29.16 | 22.8 | 22.74 | 22.03 |
| Classification of pairs | 83.63 | 81.14 | 79.88 | 79.17 |
| Replay | 65.58 | 61.22 | 64.07 | 63.89 |
| Recovery | 67.71 | 59.68 | 59.16 | 59.27 |
| STS (Semantic textual similarity) | 79.4 | 76.11 | 74.8 | 71.68 |
| MTEB (Eng, V2) | 73.3 | 69.53 | 66.01 | 66.43 |
| MTEB (Code, V1) | 76 | 65.4 | 51.94 | 58.95 |
| Xor-retaive | 90.42 | 65.67 | – | 68.76 |
| Xtreme-up | 64.33 | 34.97 | – | 18.80 |
Practical applications
- Semantic research and recovery: Improvement of the document and the corresponding passage through languages
- Classification and clustering: Categorization of robust text and grouping of documents
- Generation with recovery (CLOTH)): Improved recovery precision for applications supported by LLM
- Transversal and multilingual applications: Effortless management of internationalized content
Integration and ecosystem
- API access: Use Gemini-Embedding-001 in the Gemini API, Google AI Studio and Vertex AI.
- Transparent integration: Compatible with the main vector database solutions and AI-based AI platforms, allowing easy deployment in modern pipelines and data applications.
Price and migration
| Floor | Price | Notes |
|---|---|---|
| Free | Limited use | Ideal for prototyping and experimentation |
| Paid | $ 0.15 per 1 million tokens | Scales for production needs |
- Depreciation calendar::
gemini-embedding-exp-03-07: Not recommended on August 14, 2025- Previous models (Embedding-001, Text-Embedding-004): depreciation until the beginning of 2026
- Migration: It is recommended to Migrate to Gemini-Embedding-001 To benefit from continuous improvements and support.
Look forward to
- Batching: Google has announced the upcoming support for prizes to activate Asynchronous and profitable integration generation large -scale.
- Multimodal integrations: Future updates can allow unified interests for not only text but also code and images, progressing the extent of Gemini applications.
Conclusion
The general availability of Gemini-Embedding-001 marks a major progression in the AI toolbox of Google, offering developers a powerful, flexible and multilingual text integration solution that adapts to a wide range of application needs. With its evolutionary dimensionality, its high -level multilingual performance and its transparent integration in the research ecosystems of Popular AI and Vector, this team team model to create smarter, faster and more relevant applications on a global scale. While Google continues to innovate with features such as prize treatment and multimodal support, Gemini-Embedding-001 sets a solid base for the future of semantic understanding in AI.
| Discover the Technical details. All the merit of this research goes to researchers in this project. Ready to connect with 1 million developers / engineers / researchers? Find out how NVIDIA, LG AI Research and the best IA companies operate Marktechpost to reach their target audience (Learn more) |
Asif Razzaq is the CEO of Marktechpost Media Inc .. as a visionary entrepreneur and engineer, AIF undertakes to exploit the potential of artificial intelligence for social good. His most recent company is the launch of an artificial intelligence media platform, Marktechpost, which stands out from its in-depth coverage of automatic learning and in-depth learning news which are both technically solid and easily understandable by a large audience. The platform has more than 2 million monthly views, illustrating its popularity with the public.
