Models

OpenSploit supports 75+ LLM providers through the AI SDK. Configure providers and select models to power your security assessments.

Provider Setup

Before using a model, configure the provider via /connect:

/connect

Select your provider and enter your API key.

Model Selection

Select a model with the /models command:

/models

Or use the keybind ctrl+x m to open the model selector.

Recommended Models

For security testing, we recommend models with strong reasoning and tool-calling capabilities:

| Model | Provider | Best For | |-------|----------|----------| | Claude Opus 4.5 | Anthropic | Complex reasoning, exploitation | | Claude Sonnet 4 | Anthropic | Balanced performance | | GPT-4o | OpenAI | General purpose | | Gemini Pro | Google | Fast enumeration |

Default Model

Set a default model in opensploit.json:

{
  "model": "anthropic/claude-sonnet-4-20250514"
}

Format: provider_id/model_id

Model Configuration

Configure model-specific settings:

{
  "model": "anthropic/claude-sonnet-4-20250514",
  "models": {
    "anthropic/claude-sonnet-4-20250514": {
      "temperature": 0.7,
      "maxTokens": 8192
    }
  }
}

Variants

Variants are different configurations for the same model. Use them for different tasks:

{
  "models": {
    "anthropic/claude-sonnet-4-20250514": {
      "variants": {
        "default": {
          "temperature": 0.7
        },
        "precise": {
          "temperature": 0.2
        },
        "creative": {
          "temperature": 0.9
        }
      }
    }
  }
}

Cycle between variants with ctrl+t.

Local Models

Run models locally with Ollama:

1. Install Ollama:

   curl -fsSL https://ollama.ai/install.sh | sh
   

2. Pull a model:

   ollama pull llama3.2
   ollama pull codellama
   

3. Connect in OpenSploit:

   /connect
   

   Select "Ollama" and configure.

4. Set as default:

   {
     "model": "ollama/llama3.2"
   }
   

Model Priority

At startup, OpenSploit selects models in this order:

1. Command-line flag (--model or -m)
2. Config file specification
3. Last used model
4. First available model

Per-Agent Models

Assign different models to different agents:

{
  "agents": {
    "recon": {
      "model": "anthropic/claude-haiku-3"
    },
    "exploit": {
      "model": "anthropic/claude-opus-4"
    }
  }
}

Use faster models for simple tasks, powerful models for complex reasoning.

Token Limits

Be aware of context limits when working with large scan outputs:

| Model | Context Window | |-------|----------------| | Claude Opus 4.5 | 200K tokens | | Claude Sonnet 4 | 200K tokens | | GPT-4o | 128K tokens | | Llama 3.2 | 128K tokens |

Use /compact to summarize long sessions and free up context.