diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
index 97d04a03a78..854bc9bbb81 100644
--- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
+++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
@@ -108,6 +108,74 @@ def test_generic_cost_per_token_prefers_audio_per_second_rate() -> None:
     assert prompt_cost == pytest.approx(2 * 0.00016)
 
 
+def test_generic_cost_per_token_prefers_image_per_image_rate() -> None:
+    model_info: ModelInfo = {
+        "key": "gemini-embedding-2",
+        "max_tokens": None,
+        "max_input_tokens": None,
+        "max_output_tokens": None,
+        "input_cost_per_token": 2e-7,
+        "input_cost_per_image_token": 4.5e-7,
+        "input_cost_per_image": 0.00012,
+        "output_cost_per_token": 0.0,
+        "litellm_provider": "vertex_ai",
+        "mode": "embedding",
+        "supported_openai_params": None,
+    }
+    usage = Usage(
+        prompt_tokens=258,
+        completion_tokens=0,
+        total_tokens=258,
+        prompt_tokens_details=PromptTokensDetailsWrapper(
+            image_tokens=258,
+            image_count=1,
+        ),
+    )
+
+    prompt_cost, _ = generic_cost_per_token(
+        model="gemini-embedding-2",
+        usage=usage,
+        custom_llm_provider="vertex_ai",
+        model_info=model_info,
+    )
+
+    assert prompt_cost == pytest.approx(0.00012)
+
+
+def test_generic_cost_per_token_prefers_video_per_second_rate() -> None:
+    model_info: ModelInfo = {
+        "key": "gemini-embedding-2",
+        "max_tokens": None,
+        "max_input_tokens": None,
+        "max_output_tokens": None,
+        "input_cost_per_token": 2e-7,
+        "input_cost_per_video_token": 1.2e-5,
+        "input_cost_per_video_per_second": 0.00079,
+        "output_cost_per_token": 0.0,
+        "litellm_provider": "vertex_ai",
+        "mode": "embedding",
+        "supported_openai_params": None,
+    }
+    usage = Usage(
+        prompt_tokens=516,
+        completion_tokens=0,
+        total_tokens=516,
+        prompt_tokens_details=PromptTokensDetailsWrapper(
+            video_tokens=516,
+            video_length_seconds=2,
+        ),
+    )
+
+    prompt_cost, _ = generic_cost_per_token(
+        model="gemini-embedding-2",
+        usage=usage,
+        custom_llm_provider="vertex_ai",
+        model_info=model_info,
+    )
+
+    assert prompt_cost == pytest.approx(2 * 0.00079)
+
+
 def test_missing_cache_read_uses_off_peak_input_rate():
     from datetime import datetime, timezone
 
diff --git a/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py b/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py
index 926570d7929..fd8c2a9cf6a 100644
--- a/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py
+++ b/tests/test_litellm/llms/vertex_ai/gemini_embeddings/test_batch_embed_content_transformation.py
@@ -546,6 +546,45 @@ class TestProcessEmbedContentResponseUsage:
         )
         assert prompt_cost == pytest.approx(258 * 4.5e-7)
 
+    @pytest.mark.parametrize(
+        "input_value,resolved_files,expected_image_tokens",
+        [
+            (GCS_URL, {}, 258),
+            ("gs://my-bucket/clip.mp4", {}, 0),
+            ("gs://my-bucket/unknown.bin", {}, 0),
+            ("files/image-123", {"files/image-123": {"mime_type": "image/jpeg"}}, 258),
+            ("files/missing", {}, 0),
+            ("data:application/octet-stream;base64,abc", {}, 0),
+            ([[IMAGE_DATA_URI]], {}, 258),
+            ([], {}, 0),
+        ],
+    )
+    def test_missing_modality_details_classifies_image_inputs(self, input_value, resolved_files, expected_image_tokens):
+        response_json = {
+            "embedding": {"values": [0.1]},
+            "usageMetadata": {
+                "promptTokenCount": 258,
+                "totalTokenCount": 258,
+            },
+        }
+        result = process_embed_content_response(
+            input=input_value,
+            model_response=EmbeddingResponse(),
+            model=self.MODEL,
+            response_json=response_json,
+            resolved_files=resolved_files,
+        )
+        assert result.usage.prompt_tokens_details.image_tokens == expected_image_tokens
+        assert result.usage.prompt_tokens_details.text_tokens == 0
+
+        prompt_cost, _ = generic_cost_per_token(
+            model=self.MODEL,
+            usage=result.usage,
+            custom_llm_provider="vertex_ai",
+        )
+        expected_rate = 4.5e-7 if expected_image_tokens else 2e-7
+        assert prompt_cost == pytest.approx(258 * expected_rate)
+
     def test_mixed_text_and_image_without_modality_details_not_billed_as_image(self):
         response_json = {
             "embedding": {"values": [0.1]},