The Glean model hub enables secure access to a diverse array of leading commercial and open-sourced large language models (LLMs). Users can pick-and-choose models for their agents and each step in an agentic workflow without needing technical expertise or engineering resources. The model hub currently supports over 15+ models from providers like OpenAI, Google, Anthropic, Amazon, and DeepSeek. New models will be regularly added to the model hub as they become available on leading cloud providers Amazon Bedrock, Google Vertex AI, and Azure OpenAI. Customers have the option to bring their own model key, or take advantage of Glean’s universal model key—which unlocks access to over 15+ models pre-procured and all managed by Glean in one place. The universal model key includes agreements with model hosting providers to ensure customer enterprise data is never retained or used for model training.

​
Supported models

The following models are supported in Glean Agents:

  • Open AI models via Azure or OpenAI:
    • GPT 4.1
    • GPT 4.1 mini
    • GPT-4o
    • GPT-4o-mini
    • o3
    • o3-mini
    • o4-mini
    • o1 (Supported on customer key only)
  • Google Gemini via Vertex AI:
    • Gemini 2.5 Pro Preview (temporarily supported, will be replaced by GA model when it becomes available)
    • Gemini 2.0 Flash
    • Gemini 1.5 Pro
    • Gemini 1.5 Flash
  • Anthropic models via Google Vertex AI or Amazon Bedrock:
    • Claude 3.7 Sonnet
    • Claude 3.5 Sonnet V2
    • Claude 3.5 Haiku
  • Amazon Nova Pro 1.0 via Amazon Bedrock
  • DeepSeek R-1 via Amazon Bedrock

Note: If your deployment is glean hosted, but the Glean Assistant is configured as customer key, models via Vertex AI will not be available.

​
Enabling model hub for Glean universal model key

With the Glean universal model key, models are preconfigured, the capacity is managed, and the admin just has the ability to enable or disable models. Additionally when models are deprecated by the model providers, they will be automatically replaced with the latest recommended version by the provider.

To enable model hub for agents from a hosting provider that is different from the provider selected for the Assistant, follow these steps on the admin console

  1. Select LLM under the platform tab.
  2. Click on the “Add LLM” button
  3. Select a hosting provider.
  4. Enable all models available on the hosting provider
  5. Once enabled, the models can be disabled from the same page.

Note: At this time if GCP is your hosting provider, you cannot add models from AWS and vice versa.

​
Enabling model hub for customer keys

To enable model hub for customer keys, customers that choose to bring their own keys, follow these steps in the admin console:

  1. Select LLM under the Platform tab.
  2. Click on the “Add LLM” button
  3. Select a hosting provider.
  4. Follow the configuration steps on the page per hosting provider:
    1. Azure OpenAI
      1. Go to “Keys and Endpoints” in your Azure Portal and enter the information for Key and Endpoint to be used with Glean.
      2. Select the modes that will be used for the Assistant.
        Note: Assistant model needs to be configured as that is the default model that will be used with Agents if agent builders do not configure the Agent model in the Agent setup. If you have assistant model configured from another provider, you can skip this step
        Currently the following combinations are enabled
        1. GPT-4.1 as the large model and GPT-4.1 mini as the small model (Preferred)
        2. GPT-4o (2024-05-13) as the large model and GPT-4o-mini (2024-07-18)
      3. Select additional models that you would like to enable for Agents
      4. For each model selected, add the deployment name configured on Azure
      5. Validate to ensure that Glean can connect to the models
      6. Once validated click Save
      7. To remove a model from model choice for agents, you can go back to the same page, uncheck the model, validate the configuration and save.
    2. Amazon Bedrock - this option is only supported if you are self-hosted on AWS
      1. Enter your preferred region for the models
      2. Select the models that will be used for the Assistant.
        Note: Assistant model needs to be configured as the default model that will be used with Agents if agent builders do not configure the Agent models in the Agent setup. If you have assistant model configured from another provider, you can skip this step
        Currently the following combinations are enabled
        1. Claude 3.7 Sonnet as the large model and Claude Haiku 3.5 as the small model (Preferred)
        2. Claude 3.5 Sonnet V2 as the large model and Claude Haiku 3.5 as the small model
      3. Select additional models that you would like to enable for Agents
      4. Validate to ensure that Glean can connect to the models
      5. Once validated click Save
      6. To remove a model from model choice for agents, you can go back to the same page, uncheck the model, validate the configuration and save.
    3. Google Vertex AI - this option is only supported if you are self-hosted on GCP

      1. Select the models that will be used with the Assistant.
        Note: Assistant model needs to be configured as the default model that will be used with Agents if agent builders do not configure the Agent models in the Agent setup. If you have assistant model configured from another provider, you can skip this step

        Currently the following combinations are enabled

        1. Claude 3.7 Sonnet as the large model and Claude Haiku 3.5 as the small model (Preferred)
        2. Claude 3.5 Sonnet V2 as the large model and Claude Haiku 3.5 as the small model
        3. Gemini 1.5 Pro as the large model and Gemini 1.5 Flash as the small model
          Note: Gemini 1.5 Pro and Gemini 1.5 Flash will be replaced by Gemini 2.0 and 2.5 models shortly.

      2. Select additional models that you would like to enable for Agents

      3. Ensure that the models selected are enabled in the Google Model Garden

      4. Validate to ensure that Glean can connect to the models

      5. Once validated click Save

      6. To remove a model from model choice for agents, you can go back to the same page, uncheck the model, validate the configuration and save.

    4. OpenAI

      1. Enter the API key configured in your OpenAI Portal and enter the information below.

      2. Select the models that will be used with the Assistant. Currently the following combinations are enabled

        1. GPT-4.1 as the large model and GPT-4.1 mini as the small model (Preferred)
        2. GPT-4o (2024-05-13) as the large model and GPT-4o-mini (2024-07-18)

      3. Select additional models that you would like to enable for Agents

      4. Validate to ensure that Glean can connect to the models

      5. Once validated click SaveTo remove a model from model choice for agents, you can go back to the same page, uncheck the model, validate the configuration and save.

Note: At this time if GCP is your hosting provider, you cannot add models from AWS and vice versa.

​
How to change the model used by your agent

To set the default model that will be used in all steps of the agent, go to agent settings (gear icon in the header) and click on Select Model.

​
How to change the model used in an agent step

You also have the option of changing the model for only a single step.

Click on the step in the canvas and then click on the three dots menu in the drawer and select Advanced settings.

Then click on Select Model.