From 61096af29c0bc7b7bd34d93274a237554478522a Mon Sep 17 00:00:00 2001
From: Akhilesh Kumar <akhilbussiness@gmail.com>
Date: Wed, 15 Apr 2026 01:57:31 +0000
Subject: [PATCH] docs: Add Gemma 4 routing instructions to README and docs

---
 README.md                  |  3 +++
 docs/cli/gemma4-routing.md | 48 ++++++++++++++++++++++++++++++++++++++
 docs/sidebar.json          |  4 ++++
 3 files changed, 55 insertions(+)
 create mode 100644 docs/cli/gemma4-routing.md

diff --git a/README.md b/README.md
index 10458b2126..c5d614c693 100644
--- a/README.md
+++ b/README.md
@@ -16,6 +16,9 @@ Learn all about Gemini CLI in our [documentation](https://geminicli.com/docs/).
 
 ## 🚀 Why Gemini CLI?
 
+- **🚀 Gemma 4 Support**: Route requests to Gemma 4 models (`gemma-4-26b-a4b-it`
+  or `gemma-4-31b-it`) using the `model.gemma4Variant` configuration setting.
+
 - **🎯 Free tier**: 60 requests/min and 1,000 requests/day with personal Google
   account.
 - **🧠 Powerful Gemini 3 models**: Access to improved reasoning and 1M token
diff --git a/docs/cli/gemma4-routing.md b/docs/cli/gemma4-routing.md
new file mode 100644
index 0000000000..eede0126ea
--- /dev/null
+++ b/docs/cli/gemma4-routing.md
@@ -0,0 +1,48 @@
+# Gemma 4 Routing
+
+Gemini CLI allows you to seamlessly route your requests to Gemma 4 models. When
+enabled, requests that would normally be sent to standard Gemini Pro and Flash
+models are automatically redirected to your chosen Gemma 4 variant.
+
+## Configuration
+
+You can enable Gemma 4 routing using the CLI settings or by modifying your
+`settings.json` file.
+
+### Via Settings UI
+
+1. Open the settings dialog by running `/settings`.
+2. Navigate to the **Model** section.
+3. Locate the **Gemma 4 Variant** setting.
+4. Select your preferred model:
+   - `gemma-4-26b-a4b-it` (Gemma 4 26B A4B IT)
+   - `gemma-4-31b-it` (Gemma 4 31B IT)
+5. Save the settings and restart the CLI if prompted.
+
+### Via `settings.json`
+
+You can also set this directly in your `.gemini/settings.json` file:
+
+```json
+{
+  "model": {
+    "gemma4Variant": "gemma-4-31b-it"
+  }
+}
+```
+
+## How it works
+
+When a `gemma4Variant` is selected, Gemini CLI intercepts model resolution:
+
+- Requests for `gemini-pro`, `gemini-flash`, and their associated aliases (like
+  `auto`, `pro`, `flash`) are routed to the selected Gemma 4 model.
+- The **router model** (`flash-lite`) remains unaffected and continues to use
+  `gemini-2.5-flash-lite`. This ensures fast, lightweight background routing
+  tasks continue to operate optimally.
+- If you do not have preview model access, the CLI normally falls back to stable
+  models; however, the Gemma 4 variant will still take precedence for Pro and
+  Flash targets.
+
+To disable Gemma 4 routing, simply remove the `gemma4Variant` configuration from
+your settings or set it to `undefined`/empty in the UI.
diff --git a/docs/sidebar.json b/docs/sidebar.json
index ad5741699e..fc1268818f 100644
--- a/docs/sidebar.json
+++ b/docs/sidebar.json
@@ -171,6 +171,10 @@
             "label": "Project context (GEMINI.md)",
             "slug": "docs/cli/gemini-md"
           },
+          {
+            "label": "Gemma 4 Routing",
+            "slug": "docs/cli/gemma4-routing"
+          },
           { "label": "Settings", "slug": "docs/cli/settings" },
           {
             "label": "System prompt override",