> For the complete documentation index, see [llms.txt](https://docs.k3-labs.com/introduction/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.k3-labs.com/introduction/ai-automation-suite/choosing-ai-model.md).

# Choosing AI Model

#### Why Does Your AI Model Choice Matter?

When building automations in K3 Labs, the AI model you select has a direct impact on what your workflows can do - and the quality and reliability of the results. Some models are best for fast, simple tasks. Others can handle complex, multi-step operations, tool use, and troubleshooting.

***

#### Types of AI Models

**Basic & Fast Models**

* **Examples:** ChatGPT 4.0, Gemini 2.0 Flash, Gemini 2.0 Flashlight, GPT-3.5
* **Best for:** Simple, direct requests; summaries; basic data transformations
* **How they work:**\
  These models attempt to complete your prompt in a single step. They’ll run your task once—if there’s an issue (like an error from an API or database), they don’t retry or troubleshoot.
* **Limitations:**\
  Not ideal for complex logic or ambiguous instructions.

**Advanced & Reasoning Models**

* **Examples:** Claude Sonnet 4.0, Gemini 2.5 Pro, GPT-4o, Claude Opus
* **Best for:** Multi-step logic; chaining actions; using external tools; analytics; error handling
* **How they work:**\
  Advanced models can break tasks into steps, reason through processes, and troubleshoot as they go. If they encounter an error (e.g., a failed database query), they’ll try to fix it and re-run the action. They’re better at open-ended prompts and more robust in workflows that require connecting to external tools.
* **Strengths:**
  * Handle vague or multi-part requests
  * Better at working with integrations (Web3 Data Fetcher, MCP, etc.)
  * Retry and troubleshoot when something goes wrong

***

#### K3 Free AI Access

K3 Labs provides **free access to Gemini 2.0 Flash and Gemini 2.0 Flashlight models**. These models are available out-of-the-box for all users—no setup required, and no API key needed. Just select Gemini (K3 Free Credits) when configuring your AI agent or workflow step.

***

#### What Does This Mean for You?

* **Output Quality Will Vary:**\
  Results depend on the model’s abilities. For simple automations, fast/basic models are fine. For anything complex or critical, advanced models will generally perform better.
* **Tool Use Depends on Model:**\
  Advanced models handle tool use, integrations, and retries better. Basic models will do exactly what you ask, one time.
* **Try Different Models:**\
  For basic summaries or checks, start with Gemini 2.0 Flash (free). For multi-step, critical, or error-prone workflows, upgrade to an advanced model.

***

#### Example Scenario

* **Scenario:**\
  You ask your AI agent to “fetch token balances, summarize those above $1,000, and notify me of any new tokens this week.”
* **With a basic model:**\
  The agent does one pass. If there’s a failure or ambiguity, you may get a partial or failed result.
* **With an advanced model:**\
  The agent checks, retries, troubleshoots, and completes each step—even if it requires multiple attempts or decisions.

***

#### Summary Table

| Model Type           | Examples                                  | Best For                  | Tool Use | Error Handling           |
| -------------------- | ----------------------------------------- | ------------------------- | -------- | ------------------------ |
| Basic / Fast         | ChatGPT 4.0, Gemini 2.0 Flash             | Simple prompts            | Basic    | None (single shot)       |
| Advanced / Reasoning | Claude Sonnet 4.0, Gemini 2.5 Pro, GPT-4o | Multi-step, complex tasks | Full     | Retries, troubleshooting |

***

**Tip:**\
For most users, starting with Gemini 2.0 Flash (free) is the easiest way to try out AI automations. If your workflow is complex or critical, upgrade to a more advanced model for better performance and reliability.