> ## Documentation Index
> Fetch the complete documentation index at: https://docs.noxus.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Models & Providers
> Understanding AI models, providers, and model selection in Noxus
## What are Models?
Models are the AI engines that power intelligent operations in Noxus. They process text, make decisions, and generate responses based on their training and configuration. Noxus provides a unified interface to various AI models from leading providers, making it easy to leverage state-of-the-art AI capabilities.
Models serve as the foundation for all AI operations in the platform, from
simple text generation to complex multi-modal processing. Each model brings
unique capabilities and strengths, allowing you to choose the best fit for
your specific use case.
## Supported Providers
| Provider | Key Features | Capabilities |
| :------------------------------------------------------------------------------ | :------------------------------------------ | :----------------------------------------------------- |
| [OpenAI](https://openai.com/) | Full parameter control, vision capabilities | Tool integration, JSON mode, multi-modal processing |
| [Anthropic](https://www.anthropic.com/) | Advanced reasoning, extended thinking | Temperature control, context handling, safety features |
| [Google Vertex AI](https://cloud.google.com/vertex-ai) | Enterprise features, multi-region | Advanced APIs, token management, scalability |
| [Gemini](https://ai.google.dev/) | Google AI Studio, fast inference | Vision, function calling, large context windows |
| [AWS Bedrock](https://aws.amazon.com/bedrock/) | Multi-region, AWS-native | Claude, Titan, Nova models via AWS |
| [Azure OpenAI](https://azure.microsoft.com/products/ai-services/openai-service) | Enterprise Azure integration | GPT models with Azure compliance |
| [Azure AI Foundry](https://ai.azure.com/) | Claude models on Azure | Anthropic Messages API via Azure infrastructure |
| [DeepSeek](https://www.deepseek.com/) | Vision capabilities, reasoning | Tool integration, efficient processing, cost-effective |
| [Grok (xAI)](https://x.ai/) | Real-time knowledge, fast inference | Function calling, large context windows |
| [Mistral AI](https://mistral.ai/) | Vision processing, tool integration | Temperature sampling, safe mode, cost efficiency |
| [Groq](https://groq.com/) | Ultra-fast inference | Low-latency processing, cost-effective |
| [OpenRouter](https://openrouter.ai/) | Unified access to 100+ models | Open-source and proprietary models |
| [Perplexity](https://www.perplexity.ai/) | Real-time search capabilities | Web-grounded responses |
### Connecting Providers
Connecting a provider is fast and secure. Add your API key (or complete an OAuth flow where required), and Noxus handles credentials, secure storage, and API communication. Once connected, models from that provider are available across your agents and flows with a consistent interface.
**Provider Configuration**
* **Authentication Management**: Securely store API keys or OAuth credentials with encryption.
* **Health Monitoring**: Noxus continuously monitors provider connections and automatically handles failures with fallback routing.
* **Multi-Region Support**: Providers like Vertex AI and AWS Bedrock support multi-region configuration for data residency and latency optimization.
## Model Capabilities
| Capability | Description | Features |
| :------------------- | :-------------------------------------------- | :-------------------------------------------------- |
| **Text Processing** | Natural language understanding and generation | Chat completion, text generation, analysis |
| **Vision** | Image understanding and processing | Image analysis, multi-modal chat, visual QA |
| **Function Calling** | Structured tool use | API integration, data extraction, agent tools |
| **Reasoning** | Extended thinking and chain-of-thought | Complex analysis, multi-step problem solving |
| **JSON Mode** | Structured output generation | Reliable data extraction, schema-constrained output |
Model capabilities vary by provider and specific model version. The model picker shows capability badges for each model.
## Model Selection
Noxus offers two ways to select models for your flows and agents:
### Model Presets
Presets are named bundles of models tried in priority order. Instead of hardcoding a specific model, you reference a preset like **Cost-Effective Models** or **High-Performance Models** — the platform picks the best available model at runtime based on your active providers.
Presets handle fallback automatically: if the first model encounters an error, the next one in the chain is used.
### Custom Selection
For fine-grained control, switch to **Custom models** to hand-pick specific models from your connected providers. The model browser shows speed and quality gauges, capabilities, and provider information for each model.
## Model Configuration
**Core Configuration Options:**
* **Model Selection**: Choose from presets or custom models across all connected providers.
* **Temperature**: Control response creativity (0.0 for focused, 1.0 for creative)
* **Max Tokens**: Set response length limits
* And more per-node parameters depending on the node type.
## Next Steps
Configure providers and manage model availability.
Set up named model bundles for consistent selection.
Choose the right model for your task.
Extend the platform with custom integrations.