Add separate Gemma 4 README section above Gemma

Gemma 4 notebooks were grouped under the Gemma section together with
Gemma 3, Gemma 2, FunctionGemma, EmbeddingGemma and CodeGemma. Split
them out into their own "### Gemma 4 Notebooks" section so the newer
family is visually separated from the legacy Gemma models.

Two changes:

1. ARCHITECTURE_MAPPING gains a "gemma4" -> "Gemma 4" entry. Because
   _get_arch in extract_model_info_refined sorts mapping keys by
   length descending, "gemma4" (6 chars) is tried before "gemma"
   (5 chars). The word-boundary regex \\bgemma4\\b matches the word
   in "Gemma4 (E4B)" / "Gemma4 (E2B)" / "Gemma4 (31B)" /
   "Gemma4 (26B A4B)" but does NOT match Gemma3, Gemma3N, Gemma2,
   FunctionGemma, EmbeddingGemma or CodeGemma -- so only Gemma 4
   notebooks move into the new section, and everything else stays
   in the existing Gemma section.

2. update_readme reorders list_models so "Gemma 4" appears
   immediately before "Gemma". Default alphabetical sort would have
   placed it after Gemma (because "Gemma" is a prefix of "Gemma 4"),
   but the user wants the newer family on top. Surgical reorder via
   list.remove + list.insert at the index of "Gemma".

Result: README now has

  ### Gemma 4 Notebooks      <- 10 rows (E4B / E2B / 31B / 26B A4B
                                 x Vision / Conversational / Audio)
  ### Gemma Notebooks        <- 17 rows (Gemma3N, Gemma3, Gemma2,
                                 FunctionGemma, EmbeddingGemma,
                                 CodeGemma)

Author: danielhanchen

README.md

@@ -155,7 +155,7 @@ Below are Colab notebooks, organized by model. You can also view all [notebooks
 | **(DGX Spark) gpt oss** **(20B)** | GRPO 2048 | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt_oss_(20B)_Reinforcement_Learning_2048_Game_DGX_Spark.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 | **(A100) gpt oss** **(120B)** | Fine Tuning | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt-oss-(120B)_A100-Fine-tuning.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 
-### Gemma Notebooks
+### Gemma 4 Notebooks
 | Model | Type | Notebook Link |
 | --- | --- | --- |
 | **Gemma4** **(E4B)** | Vision | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma4_(E4B)-Vision.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
@@ -168,6 +168,10 @@ Below are Colab notebooks, organized by model. You can also view all [notebooks
 | **Gemma4** **(31B)** | Conversational | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma4_(31B)-Text.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 | **Gemma4** **(26B A4B)** | Vision | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma4_(26B_A4B)-Vision.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 | **Gemma4** **(26B A4B)** | Conversational | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma4_(26B_A4B)-Text.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
+
+### Gemma Notebooks
+| Model | Type | Notebook Link |
+| --- | --- | --- |
 | **Gemma3N** **(4B)** | Vision | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N_(4B)-Vision.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 | **Gemma3N** **(4B)** | Multimodal | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N_(4B)-Conversational.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |
 | **Gemma3N** **(4B)** | Audio | <a href="https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N_(4B)-Audio.ipynb" target="_blank" rel="noopener noreferrer"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |

update_all_notebooks.py

@@ -805,6 +805,11 @@ def _replace(m):
 
 ARCHITECTURE_MAPPING = {
     # Gemma Family
+    # NOTE: "gemma4" must appear before "gemma" so that the longest-key-first
+    # match in extract_model_info_refined routes Gemma 4 notebooks to their
+    # own section. Other Gemma* notebooks (Gemma3, Gemma3N, Gemma2,
+    # FunctionGemma, EmbeddingGemma, CodeGemma) still resolve to "Gemma".
+    "gemma4": "Gemma 4",
     "gemma": "Gemma",
     "codegemma": "Gemma", # Explicitly map specific models if needed
 
@@ -2950,6 +2955,13 @@ def update_readme(
         if arch not in list_models:
             list_models.append(arch)
 
+    # Place "Gemma 4" immediately before "Gemma" so the newer family appears
+    # above its predecessor instead of after it (default alphabetical sort
+    # would order "Gemma" before "Gemma 4" because "Gemma" is a prefix).
+    if "Gemma 4" in list_models and "Gemma" in list_models:
+        list_models.remove("Gemma 4")
+        list_models.insert(list_models.index("Gemma"), "Gemma 4")
+
     # Cross-cutting sections (notebooks can appear in multiple sections)
     for cross_section in ["Vision (Multimodal)", "Embedding", "OCR"]:
         if cross_section not in list_models: