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

# Models

> Use the free tier, or bring any OpenAI-compatible model. Recommended models and exact settings.

The [free tier](/free-tier) runs on a managed model and is on by default. When you want a specific model or heavier usage, bring your own: Adsum ships a curated provider picker, the free tier, the GLM Coding Plan, Claude, DeepSeek V4, and any OpenAI or Anthropic-compatible endpoint, cloud or local.

<Frame caption="The provider picker: the free tier, your own coding agent (beta), the GLM Coding Plan, Claude, and any OpenAI- or Anthropic-compatible or local endpoint.">
  <img src="https://mintcdn.com/adsumnetworks/mGdITdC6ye-hfbtR/images/provider-picker.png?fit=max&auto=format&n=mGdITdC6ye-hfbtR&q=85&s=a641b9eadeed56804c75ad07c84cca97" alt="Adsum IoT Coder API provider picker showing Adsum Free Tier, Your own coding agent (beta), GLM Coding Plan, Anthropic (Claude), OpenRouter, OpenAI Compatible, and Anthropic Compatible" width="875" height="520" data-path="images/provider-picker.png" />
</Frame>

<Note>
  The model must have strong native **tool-calling** (function-calling). Models without it cannot execute hardware actions or debug workflows.
</Note>

## Recommended models

| Model                          | Best for                     | Notes                                                                                                                                                       |
| :----------------------------- | :--------------------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Claude Sonnet 5**            | The strongest results        | The default in the picker. Best reasoning and tool use for hard, multi-step firmware work.                                                                  |
| **Claude Haiku 4.5**           | Speed and cost               | The model our [benchmark](/benchmark) ran on. Fast and inexpensive for everyday loops.                                                                      |
| **GLM Coding Plan**            | Cost-effective, long context | A flat z.ai subscription: `glm-5.2` (1M context), plus `glm-5-turbo` and `glm-4.7`. Tool-calling optimized.                                                 |
| **DeepSeek V4** (flash / pro)  | Cost-sensitive long sessions | Large context for long debug loops at a low price per token. Via [OpenRouter](https://openrouter.ai) or the [DeepSeek API](https://api-docs.deepseek.com/). |
| **Local** (Ollama / LM Studio) | Everything on your machine   | Keeps code and logs on your machine. Needs strong native tool-calling.                                                                                      |

<Frame caption="Configuring the GLM Coding Plan: paste your z.ai coding-plan key, and the endpoint and model list are set for you.">
  <img src="https://mintcdn.com/adsumnetworks/JgCemrbEGYYNCeWH/images/glm-config.png?fit=max&auto=format&n=JgCemrbEGYYNCeWH&q=85&s=869eb129332d8599ff4c6b82973eab28" alt="Adsum IoT Coder settings: GLM Coding Plan provider with API key, model glm-5-turbo, an extended-thinking toggle, and 200K context" width="875" height="935" data-path="images/glm-config.png" />
</Frame>

## Configure a provider

Open the Adsum settings, choose a provider, and add your key. Providers include the GLM Coding Plan, Anthropic (Claude), OpenRouter, OpenAI-Compatible, Anthropic-Compatible, and a local Ollama or LM Studio server.

**Recommended setup for Claude Haiku 4.5**, matching the [benchmark](/benchmark) configuration:

| Setting                   | Value                                          |
| :------------------------ | :--------------------------------------------- |
| API provider              | OpenRouter (or any OpenAI-compatible endpoint) |
| Model                     | `anthropic/claude-haiku-4.5`                   |
| Enable thinking           | Off                                            |
| Prompt caching (Advanced) | On                                             |

## Provider quick setup

The endpoint URL is the only provider-specific part. Use these as the API endpoint in the extension settings, then add your key.

<AccordionGroup>
  <Accordion title="GLM Coding Plan">
    Pick **GLM Coding Plan** as the provider and paste your z.ai coding-plan key. The coding endpoint and the model list (`glm-5.2`, `glm-5-turbo`, `glm-4.7`) are set for you.
  </Accordion>

  <Accordion title="OpenRouter">
    Endpoint: `https://openrouter.ai/api/v1`. One key reaches Claude, DeepSeek, GLM, and more. Set the model to the provider-prefixed id, for example `anthropic/claude-haiku-4.5`.
  </Accordion>

  <Accordion title="DeepSeek API">
    Endpoint: `https://api.deepseek.com`. Use a DeepSeek key directly for cost-sensitive long sessions.
  </Accordion>

  <Accordion title="Local (Ollama or LM Studio)">
    Run a model on your own machine and point Adsum at it: Ollama is `http://localhost:11434/v1`, LM Studio is `http://localhost:1234/v1`. The model still needs strong tool-calling to drive hardware workflows.
  </Accordion>

  <Accordion title="Anthropic-Compatible">
    Point Adsum at any endpoint that speaks the Anthropic wire format: pick **Anthropic-Compatible**, set the base URL and model id, and add your key.
  </Accordion>
</AccordionGroup>