← All posts

4 min read GenViz Team conversation analyticsAI modelsrelease

Gemini 3.8 Flash and Gemini 4 Argon join the GenViz stable

Two Google models for conversation analytics: try Gemini 3.8 Flash or Gemini 4 Argon with your own Google Gemini API key or your organization’s OpenRouter key.

There are two new arrivals in the GenViz stable: Gemini 3.8 Flash and Gemini 4 Argon. They bring Google’s Gemini models to your conversations, with two ways to reach them: your own Google Gemini API key or your organization’s OpenRouter key.

Both work with your existing connections and analysis tools. Choose a model, ask a question, and inspect the result before moving on.

Pick Flash or Argon for your question

In the GenViz picker, Gemini 3.8 Flash is Google’s fast Gemini for everyday analysis, while Gemini 4 Argon is Google’s most capable Gemini, for hard reasoning. Argon is a very new arrival, released on October 13.

Those descriptions give you starting points to try on your own data:

  • Try Flash for an everyday question. Ask for a grouped total, a comparison between periods, or the rows behind a familiar number.
  • Try Argon for a harder investigation. Work through business definitions, examine the contributors to a change, or check a calculation that needs more than one step.

Those are starting points, not a ranking of answers. We have not published a GenViz benchmark comparing these models. Start with a question whose answer you can verify, then judge the result against your data.

Connect through Google or OpenRouter

For direct access, an administrator adds your Google Gemini API key in Settings → LLM API keys, using the Google Gemini API key field. The models appear under the new Google tab in the model picker. The access check on the LLM API keys page shows which Gemini models the key can reach, so admins see availability before anyone picks a model.

For OpenRouter access, use your organization’s OpenRouter key. The OpenRouter account must allow Google as a provider. After enabling Google, an administrator clicks Re-check access on the LLM API keys page. The models appear under the OpenRouter tab.

Gemini is bring-your-own-key only on every GenViz plan, including Enterprise. GenViz does not provide or pay for a Gemini key.

If your organization has both keys, you see both entries. They are separate on purpose: direct Google turns show a cost breakdown by input / output, while OpenRouter turns show OpenRouter’s own reported total as one figure, labeled Provider-reported total.

Set the reasoning effort for the work

Both Gemini models always reason. They start at medium reasoning effort, with low, medium, and high available in the picker. There is no none option for Gemini. Inspect the answer and adjust for your next question.

Put the new arrivals to work

Gemini keeps its line of thought across queries, from finding a support queue with a backlog to examining the tickets behind it:

Which support queue has the most unresolved tickets? Then show those tickets by age band in a table.

Follow up on that result:

In that queue, which issue types account for the tickets over 30 days old?

Open the returned table, inspect the query, and refine the dates or business definitions in the same conversation. Saved extracts keep results available for later inspection, and Python analysis can return another result for a calculation across extracts.

Both models use the same read-only database tools and connection controls as other GenViz models. Your prompts and the context used to answer them, including relevant schema and query results, go to the selected provider. With the OpenRouter route, they go to OpenRouter as well as the model provider. Choose the route that fits your organization’s data requirements; our privacy policy explains the services involved.

Try Gemini in your next conversation

Choose Google → Gemini 3.8 Flash or Google → Gemini 4 Argon, or use OpenRouter → Gemini 3.8 Flash or OpenRouter → Gemini 4 Argon. Bring a familiar question, inspect the answer, and ask the follow-up. The stable has grown; your data is still the place to judge the result.

Explore conversation analytics or download GenViz to get started.