> ## Documentation Index
> Fetch the complete documentation index at: https://docs.neo.projectdiscovery.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> Choose which AI model powers your Neo tasks

Neo supports a range of frontier models from Anthropic, OpenAI, Google, xAI, and others. You can set a preferred model under **Settings → Models**, and override it per task or per automation.

<Frame>
  <iframe className="w-full aspect-video rounded-xl" src="https://www.youtube.com/embed/qDD2DjhJLgY" title="YouTube video player" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowFullScreen />
</Frame>

The model you choose applies to the primary reasoning agent. Sub-agents in Thorough mode may use different models optimized for their specific role.

## Choosing a model

**Auto** is the default. It selects a cost-effective model with the right intelligence level for each task automatically. For most users, Auto is the right starting point.

If you want direct control, select a specific model. Use the descriptions below to match the model to your use case.

You can also set the reasoning effort and speed for a task when you start it. Each model shows the options it supports. If you pick an option the model doesn't support, Neo lets you know before the task starts.

## Available models

### Anthropic

<div className="model-table">
  | Model | Best for |
  | - | - |
  | **Claude Opus 5** | The newest and most capable model for complex security tasks |
  | **Claude Opus 4.8** | Highly capable Opus for complex, multi-step security tasks |
  | **Claude Opus 4.7** | Previous Opus generation for complex security tasks |
  | **Claude Opus 4.6** | Slightly less capable than Opus 4.7 |
  | **Claude Sonnet 4.6** | Latest Sonnet with adaptive thinking |
  | **Claude Haiku 4.5** | Fast and efficient for quick tasks |
</div>

### OpenAI

<div className="model-table">
  | Model | Best for |
  | - | - |
  | **GPT-5.6 Sol High** | Flagship GPT-5.6 model with high reasoning for complex agentic work |
  | **GPT-5.6 Terra High** | Balanced GPT-5.6 model with high reasoning at lower cost |
</div>

### Google

<div className="model-table">
  | Model | Best for |
  | - | - |
  | **Gemini 3.8 Flash** | Fast and efficient for quick research |
</div>

### xAI

<div className="model-table">
  | Model | Best for |
  | - | - |
  | **Grok 4.6** | Latest Grok generation for frontier reasoning |
  | **Grok 4.5 High** | Frontier reasoning at high effort |
  | **Grok 4.5 Fast** | Same model, low reasoning effort for speed |
  | **Grok 4.3 High** | Frontier reasoning at high effort |
  | **Grok 4.3 Fast** | Same model, low reasoning effort for speed |
</div>

### Others

<div className="model-table">
  | Model | Best for |
  | - | - |
  | **Kimi K3** | 2.8T multimodal reasoning for coding and agentic workflows |
  | **GLM 5.3** | Z.ai flagship with strong agentic reasoning |
  | **GLM 5.3 Flash** | Fast, efficient variant of GLM 5.3 for quicker tasks |
  | **DeepSeek V4 Pro High** | Frontier open reasoning at high effort |
  | **DeepSeek V4.1 Flash** | Newest DeepSeek Flash with a 1M context window for very large codebases |
  | **DeepSeek V4 Flash 0731** | Fast, efficient reasoning with a 1M context window |
</div>

## Overriding the model

You can override the default model for a specific automation in the automation's **Configuration** section. This lets you run different automations on different models without changing your global preference.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.