Models

Huginn uses a single configurable model for all activity. Each named agent can override the default with its own model, so your team can still run different models for different roles.

Default model

Set the global default in ~/.huginn/config.json:

{
  "default_model": "qwen2.5-coder:14b"
}

All agents fall back to this model unless they have their own override.

Per-agent model override

When you create or configure a named agent, you can assign it a specific model:

/agents swap Steve qwen2.5-coder:32b

Or set it at creation time. The agent’s model persists across sessions in ~/.huginn/agents.json.

Switching at runtime

Use /switch-model inside Huginn, or natural language:

use deepseek-r1 for this session

Supported providers

Local (managed runtime or Ollama)

Recommended models by hardware:

RAM	Model
8 GB	`qwen2.5-coder:7b`
16 GB	`qwen2.5-coder:14b`
32 GB+	`qwen2.5-coder:32b` or `qwen3-coder:30b`

Anthropic

{
  "backend": {
    "type": "external",
    "provider": "anthropic",
    "endpoint": "https://api.anthropic.com",
    "api_key": "$ANTHROPIC_API_KEY"
  },
  "default_model": "claude-sonnet-4-6"
}

OpenRouter

Access 200+ models from a single API key:

{
  "backend": {
    "type": "external",
    "provider": "openrouter",
    "api_key": "$OPENROUTER_API_KEY"
  },
  "default_model": "anthropic/claude-sonnet-4-6"
}

You can mix providers per-agent — e.g. use a cloud model for one specialist and a local model for another.

OpenAI

{
  "backend": {
    "type": "external",
    "provider": "openai",
    "endpoint": "https://api.openai.com/v1",
    "api_key": "$OPENAI_API_KEY"
  },
  "default_model": "gpt-4o"
}