> For the complete documentation index, see [llms.txt](https://ultrasafe.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://ultrasafe.gitbook.io/docs/readme.md).

# Welcome

### Welcome to UltraSafe AI Documentation

UltraSafe AI is a leading AI research and development company, creating advanced AI models for a wide range of applications. Our platform enables developers and enterprises to build innovative products and applications powered by UltraSafe AI's state-of-the-art models.

### UltraSafe AI Large Language Models (LLMs)

We offer a variety of models, from general-purpose to highly specialized, driving innovation and convenience for our developer community. Our models excel in multilingual capabilities, code generation, mathematics, and advanced reasoning.

#### State-of-the-art generalist models

* **EUS1**: Our flagship auto-regressive language model, based on Autoregressive Transformer Architecture with expert routing capabilities. Last updated on 23 July 2024.
* **EUS Mega**: An advanced multimodal AI model for complex tasks with extensive context understanding.
* **EUS Flash**: A high-speed text model for rapid processing, efficient for small-scale applications.

#### Specialized models

**UltraSafe AI offers a range of specialized models to cater to specific industry needs:**

* UltraSafe AI - Healthcare
* UltraSafe AI - Finance
* UltraSafe AI - Code
* UltraSafe AI - Conversational
* UltraSafe AI - Education
* UltraSafe AI - Media
* UltraSafe AI - Legal
* UltraSafe AI - Cyber Security
* UltraSafe AI - Translation Engine
* UltraSafe Image Generation
* UltraSafe Document Analyzer
* UltraSafe Data Reranker

Learn more about our models here.

### Explore the UltraSafe AI APIs

The UltraSafe AI APIs empower LLM applications via:

* **Text generation**: Enables streaming and provides the ability to display partial model results in real-time.
* **Code generation**: Empowers code generation tasks, including fill-in-the-middle and code completion.
* **Embeddings**: Useful for RAG where it represents the meaning of text as a list of numbers.
* **Multilingual capabilities**: Understands and generates in 100+ human languages and 80+ programming languages.
* **Mathematical problem-solving**: Strong performance in reasoning and mathematical tasks.
* **Expert routing**: Token-level query routing to multiple expert models.
* **Iterative learning**: Capabilities for new tasks, styles, and languages.