Please add more Gemini/Gemma models

  • Resolved kthang91

    (@kthang91)


    3 weeks, 5 days ago

    Hi Limb AI Chatbot team,

    Thank you for the plugin. I’m currently using it on my WordPress site and it works well so far.

    I noticed that the Gemini model dropdown currently only shows a few models, such as:

    • Gemini 3 Flash Preview
    • Gemini 3.1 Pro Preview
    • Gemini 3.1 Flash Lite Preview

    Could you please add support for more Gemini API models, especially the Gemma models available through the Gemini API?

    The models I would like to use include:

    • gemma-4-31b-it
    • gemma-4-26b-a4b-it

    If possible, it would be even better if the plugin could fetch the available models dynamically from the Gemini API instead of using a fixed model list. This would help users select newer models as Google releases them, without waiting for a plugin update each time.

    My use case is a WordPress chatbot for customer support. Gemini 3.1 Flash Lite works, but the free-tier rate limit is quite low for live website chat. Some Gemma models have more suitable quotas for my use case, so it would be very useful to select them inside the plugin.

    Could you please consider adding these models or allowing users to enter a custom Gemini model ID manually?

    Thank you!

Viewing 3 replies - 1 through 3 (of 3 total)
  • Plugin Contributor Raffi Yeghiazaryan

    (@raffiyeghiazaryan)

    3 weeks, 5 days ago

    Hi @kthang91,

    Thanks for the clear feedback—we’ve noted your requests. We’ll review this and reply here with a more detailed update soon.

    Thanks again.

    Plugin Contributor Raffi Yeghiazaryan

    (@raffiyeghiazaryan)

    3 weeks, 2 days ago

    Hi @kthang91 ,

    Thank you for the suggestion! We ran extensive compatibility tests with the Gemma models to see if we could support them in the Limb AI Chatbot plugin.

    Unfortunately, we found that Gemma does not interpret system instructions the same way standard Gemini models do. Gemma models require a completely different prompt architecture, highly specific formatting tags, and strict engineering constraints to prevent them from leaking internal reasoning or behaving unpredictably.

    To ensure our plugin delivers a premium, reliable experience out of the box, we only shortlist models that consistently align with our core instruction set. Because Gemma requires custom-built infrastructure to behave properly, we cannot support it at this time.

    We appreciate your understanding as we focus on keeping the plugin stable and efficient!

    Thread Starter kthang91

    (@kthang91)

    3 weeks, 2 days ago

    Hi Raffi,

    Thanks for the clarification. I understand the limitations and the reasoning behind not supporting Gemma models at this time.

    Appreciate the transparency and the work you’re doing to keep the plugin stable.

    Best regards.

Viewing 3 replies - 1 through 3 (of 3 total)

You must be logged in to reply to this topic.