Using Ollama Cloud Models in OpenClaw: GLM-5, Kimi K2.5, and More

A common question in the OpenClaw Discord: can you use Ollama's cloud-hosted models like glm-5:cloud instead of running models locally? The answer is yes — but with some important configuration details.

What Are Ollama Cloud Models?

Ollama recently introduced cloud-hosted versions of certain models, identified by the :cloud suffix (e.g., glm-5:cloud, kimi-k2.5:cloud). These run on Ollama's infrastructure, so you get the benefits of larger models without needing local GPU resources.

The Configuration Challenge

The main gotcha: Ollama cloud models don't use the standard local Ollama API endpoint. They require Ollama's cloud API endpoint and authentication.

If you try to use ollama/glm-5:cloud with your local Ollama config pointing at http://localhost:11434, you'll get connection errors because the cloud model doesn't exist locally.

How to Configure Ollama Cloud Models

Option 1: Ollama's Cloud API (When Available)

If Ollama provides a cloud API endpoint with authentication:

"models": {
  "mode": "merge",
  "providers": {
    "ollama-cloud": {
      "baseUrl": "https://api.ollama.ai/v1",
      "apiKey": "your-ollama-api-key",
      "api": "openai-completions",
      "models": [
        {
          "id": "glm-5:cloud",
          "name": "GLM-5 Cloud"
        }
      ]
    }
  }
}

Then reference as ollama-cloud/glm-5:cloud.

Option 2: Use the Original Provider's API

Many Ollama cloud models are actually repackaged versions of models available directly from their creators. Often, the better approach is to use the original provider:

For GLM-5 (Zhipu AI):

"providers": {
  "zhipu": {
    "baseUrl": "https://open.bigmodel.cn/api/paas/v4",
    "apiKey": "your-zhipu-api-key",
    "api": "openai-completions",
    "models": [
      {
        "id": "glm-4-flash",
        "name": "GLM-4 Flash"
      }
    ]
  }
}

For Kimi (Moonshot AI via NVIDIA NIM):

"providers": {
  "nvidia": {
    "baseUrl": "https://integrate.api.nvidia.com/v1",
    "apiKey": "your-nvidia-api-key",
    "api": "openai-completions",
    "models": [
      {
        "id": "moonshotai/kimi-k2.5",
        "name": "Kimi K2.5",
        "reasoning": true
      }
    ]
  }
}

Common Mistakes to Avoid

1. Wrong provider prefix: ollama/glm-5:cloud assumes your local Ollama config. Use a dedicated provider for cloud models.

2. Missing API key: Cloud models require authentication, unlike local Ollama.

3. Incorrect endpoint: Cloud endpoints differ from the local http://localhost:11434 endpoint.

Quick Checklist

• ✅ Define a separate provider for cloud-hosted models
• ✅ Use the correct baseUrl for the cloud API
• ✅ Include your apiKey
• ✅ Reference the model with the correct provider prefix: provider/model-id
• ✅ Restart the gateway after config changes: openclaw gateway restart

When to Use Cloud vs Local

Need More Help?

If you're stuck, drop into the #help channel on the OpenClaw Discord with:

• Your OpenClaw version (openclaw --version)
• The specific model you're trying to use
• Any error messages from openclaw logs

The community is active and can usually help debug provider configurations quickly.

Based on discussions in the OpenClaw Discord community. Have a config that works? Share it in #help to help others!