Skip to main content
xpander provides flexible AI model configuration, allowing you to use xpander’s managed models or bring your own LLM keys and AI Gateway.

xpander Built-in Keys

Use xpander’s pre-configured models with pay-as-you-go billing

Bring Your Own Keys

Connect your own API keys from OpenAI, Anthropic, AWS, Google, etc.

AI Gateway

Route through Helicone, OpenRouter, or your custom AI Gateway

​
Supported Models & Providers

xpander integrates with a wide range of LLM providers and models. Below is a comprehensive matrix of supported providers and their available models.
Available with xpander AI built-in keys (pay-as-you-go) or bring your own OpenAI API key
  • GPT-5.2, GPT-5.1, GPT-5, GPT-5 Mini, GPT-5 Nano
  • GPT-4.1, GPT-4.1-mini
  • GPT-4o, GPT-4o Mini, GPT-4 Turbo
  • GPT-3.5 Turbo
Available with xpander AI built-in keys (pay-as-you-go) or bring your own Anthropic API key
  • Claude Opus 4.5, Claude Opus 4
  • Claude Sonnet 4.5, Claude Sonnet 4, Claude Sonnet 3.7, Claude Sonnet 3.5
  • Claude Haiku 3.5
Bring your own AWS credentials with Bedrock access
  • Claude Opus 4, Claude Sonnet 4.5, Claude Sonnet 4, Claude Sonnet 3.7, Claude Sonnet 3.5
  • Claude Haiku 3.5
  • Amazon Titan Text Express
Bring your own Google AI Studio API key
  • Gemini 3 Pro (Preview)
  • Gemini 2.5 Pro
  • Gemini 2.0 Flash, Gemini 2.0 Flash Lite
Bring your own NVIDIA API keyMeta Llama Models:
  • Llama 4 Scout 17B, Llama 4 Maverick 17B
  • Llama 3.3 70B, Llama 3.1 405B, Llama 3.1 70B, Llama 3.1 8B
  • Llama 3.2 3B, Llama 3.2 1B
Mistral Models:
  • Mistral Small 3.2 24B, Mistral 7B Instruct v0.3
NVIDIA Nemotron:
  • Nemotron Ultra 253B, Nemotron Nano 8B, Nemotron Nano 4B
Bring your own Fireworks AI API key
  • GLM-4.6
  • Kimi K2 Instruct
  • DeepSeek V3.1
  • OpenAI gpt-oss-120b, gpt-oss-20b
  • Qwen3 235B A22B (Thinking & Instruct modes)
Bring your own Helicone API keyHelicone provides access to models from multiple providers through a unified AI gateway with observability, caching, and rate limiting features.Supported Providers:
  • Anthropic Claude (Opus, Sonnet, Haiku - all versions)
  • OpenAI (GPT-5, GPT-4.1, GPT-4o, o1, o3, o4 series)
  • Google Gemini (2.5 Pro, Flash, Lite & 3 Pro Preview)
  • xAI Grok (3, 4, Code Fast)
  • Meta Llama (4, 3.3, 3.1)
Bring your own OpenRouter API keyOpenRouter provides unified access to 200+ models with automatic fallback, load balancing, and unified pricing.Frontier Models:
  • Anthropic Claude (Opus 4.5, Sonnet 4.5, Haiku 4.5)
  • OpenAI (GPT-5.1, GPT-5.1 Chat/Codex, o3/o4 Deep Research)
  • Google Gemini (3 Pro Preview, 2.5 Flash Image)
Specialized Models:
  • xAI Grok (4, 4.1 Fast)
  • DeepSeek (V3.1, V3.2)
  • Qwen3 (Max, Coder Plus, VL Thinking)
  • Amazon Nova Premier
Open Source:
  • AllenAI OLMo 3, Prime Intellect INTELLECT-3, Liquid LFM2
Bring your own Nebius API keyNebius provides high-performance inference for open-source models.Meta Llama:
  • Llama 3.3 70B, Llama 3.1 8B, Llama Guard 3 8B
NVIDIA:
  • Nemotron Ultra 253B, Nemotron Nano V2 12B
Qwen:
  • Qwen3 235B (Instruct & Thinking), Qwen3 32B
  • Qwen2.5 Coder 7B, Qwen2.5 VL 72B, Qwen3 Embedding 8B
DeepSeek:
  • DeepSeek R1, DeepSeek V3
Google Gemma:
  • Gemma 3 27B, Gemma 2 9B, Gemma 2 2B
Others:
  • Moonshot Kimi K2 (Instruct & Thinking)
  • GLM 4.5, GLM 4.5 Air
  • OpenAI gpt-oss-120b, gpt-oss-20b
Image Generation:
  • Black Forest Labs FLUX (Dev & Schnell)
Using Custom Models: The models listed above are featured models that have been tested with xpander. To use a different model from your provider, select “Custom” in the model dropdown and enter the model name exactly as specified by your provider (e.g., fireworks/custom-120b, anthropic/custom-model-id). This is particularly useful for:
  • Private or fine-tuned models only you have access to
  • Newly released models not yet in the dropdown
  • Provider-specific model variants with custom endpoints

​
How to change AI Model and Provider

Configure your AI model from the Workbench General → LLM Settings panel.
LLM Settings Overview

​
Bring Your Own Keys

You can bring your own LLM API keys and use your own AI Gateway through the configuration panel.
Custom AI Gateway Configuration
Configuration options:
  • Model Provider - Select from supported providers (OpenAI, Anthropic, Azure, AWS Bedrock, etc.)
  • API Key - Your provider’s API key for authentication
  • API Base URL - Custom AI Gateway endpoint (e.g., ai.your-company.com)
  • Model Name - Specific model version to use
If your AI Gateway is behind a private subnet or firewall, make sure to run xpander in the same network with access to those models.

​
Using the SDK

from xpander_sdk import Backend, Configuration
from agno.agent import Agent

backend = Backend(configuration=Configuration(api_key="<your-xpander-key>"))
agno_agent = Agent(**backend.get_args(agent_id="<agent-id>"))

# Agent uses the model configured in Workbench
agno_agent.print_response(input='What is xpander?')

​
How Backend Configuration Works

By default, the Backend object automatically fetches the LLM configuration from the Workbench:
  • Model Provider - The LLM provider selected in the Workbench
  • API Keys - Securely retrieved from the platform vault at runtime
  • Base URL - Custom AI Gateway endpoint (if configured)
  • Model Name - The specific model version to use
All credentials are stored securely in the vault and automatically injected when your agent runs - you never need to hardcode API keys in your code.

​
When to Override the Model

Override the model in your code only when you need to:
  • Test with different models without changing Workbench settings
  • Use different models for different agent instances
  • Switch models dynamically based on runtime conditions
  • Run local models (Ollama) for development
When you override agent.model in your code, you’re responsible for providing the API keys for that model. The platform vault credentials only apply to the default Workbench configuration.

​
How to use SDK model overrides in production

When running the agent in production, you will use the @on_task decorator to stream events to the agent. Learn more about the Backend class in the API reference
xpander_handler.py
from dotenv import load_dotenv
load_dotenv()

from xpander_sdk import Task, on_task, Backend, Tokens
from agno.agent import Agent

@on_task
async def my_agent_handler(task: Task):
    # Get xpander agent details
    backend = Backend(configuration=task.configuration)

    # Create Agno agent instance
    agno_args = await backend.aget_args(task=task)
    agno_agent = Agent(**agno_args)

    # Run the agent
    result = await agno_agent.arun(input=task.to_message(), files=task.get_files(), images=task.get_images())

    task.result = result.content

    # report execution metrics
    task.tokens = Tokens(prompt_tokens=result.metrics.input_tokens, completion_tokens=result.metrics.output_tokens)
    task.used_tools = [tool.tool_name for tool in result.tools]

    return task
When you invoke the agent from the API, the task object will come with the AI Model configured inside it. You can override the parameters in two ways: Option 1: Override the model directly after agent creation 
agno_agent = Agent(**agno_args)
agno_agent.model = Ollama('llama3:8b')
Option 2: Use the override parameter in aget_args() 
# Define custom overrides
custom_overrides = {
    'model': 'gpt-4',
    'temperature': 0.7,
    'max_tokens': 2000,
    'show_tool_calls': False
}

# Resolve arguments with overrides
agno_args = await backend.aget_args(
    task=task,
    override=custom_overrides
)
agno_agent = Agent(**agno_args)
When running your agent handler locally, you can trigger it through the Workbench UI, API, or SDK. The task object automatically includes the configured LLM client with credentials from your Workbench settings.
LLM Configuration in Task Object
This enables you to dynamically adjust model parameters based on runtime conditions, task requirements, or user preferences without hardcoding credentials.