feat(voyage): add voyage-context-4, voyage-4 family, voyage-4-nano and voyage-multimodal-3.5

docs/my-website/docs/providers/voyage.md

@@ -54,6 +54,10 @@ All models listed here https://docs.voyageai.com/embeddings/#models-and-specific
 
 | Model Name              | Function Call                                              |
 |-------------------------|------------------------------------------------------------|
+| voyage-4-large          | `embedding(model="voyage/voyage-4-large", input)`          | 
+| voyage-4                | `embedding(model="voyage/voyage-4", input)`                | 
+| voyage-4-lite           | `embedding(model="voyage/voyage-4-lite", input)`           | 
+| voyage-4-nano           | `embedding(model="voyage/voyage-4-nano", input)`           | 
 | voyage-3.5              | `embedding(model="voyage/voyage-3.5", input)`              | 
 | voyage-3.5-lite         | `embedding(model="voyage/voyage-3.5-lite", input)`         | 
 | voyage-3-large          | `embedding(model="voyage/voyage-3-large", input)`          | 
@@ -72,9 +76,11 @@ All models listed here https://docs.voyageai.com/embeddings/#models-and-specific
 | voyage-lite-01          | `embedding(model="voyage/voyage-lite-01", input)`          |
 | voyage-lite-01-instruct | `embedding(model="voyage/voyage-lite-01-instruct", input)` |
 
-## Contextual Embeddings (voyage-context-3)
+## Contextual Embeddings (voyage-context-4, voyage-context-3)
 
-VoyageAI's `voyage-context-3` model provides contextualized chunk embeddings, where each chunk is embedded with awareness of its surrounding document context. This significantly improves retrieval quality compared to standard context-agnostic embeddings.
+VoyageAI's contextualized embedding models (`voyage-context-4` and the previous-generation `voyage-context-3`) provide contextualized chunk embeddings, where each chunk is embedded with awareness of its surrounding document context. This significantly improves retrieval quality compared to standard context-agnostic embeddings.
+
+`voyage-context-4` is the current model; `voyage-context-3` remains available as a legacy option. Swap the model name in the examples below to switch between them.
 
 ### Key Benefits
 - Chunks understand their position and role within the full document
@@ -149,7 +155,12 @@ print(f"Processed {len(response.data)} documents")
 | voyage-code-3 | Code retrieval and search | 32K | $0.18 |
 | voyage-finance-2 | Financial documents | 32K | $0.12 |
 | voyage-law-2 | Legal documents | 16K | $0.12 |
-| voyage-context-3 | Contextual document embeddings | 32K | $0.18 |
+| voyage-4-large | Best overall quality | 32K | $0.12 |
+| voyage-4 | General-purpose | 32K | $0.06 |
+| voyage-4-lite | Latency-sensitive applications | 32K | $0.02 |
+| voyage-4-nano | Open-weight, free | 32K | $0.00 |
+| voyage-context-4 | Contextual document embeddings (current) | 32K/doc | $0.12 |
+| voyage-context-3 | Contextual document embeddings (legacy) | 32K/doc | $0.18 |
 
 ## Rerank
 

litellm/llms/voyage/embedding/transformation_contextual.py

@@ -106,11 +106,35 @@ def transform_embedding_request(
         headers: dict,
     ) -> dict:
         return {
-            "inputs": input,
+            "inputs": self._normalize_inputs(input),
             "model": model,
             **optional_params,
         }
 
+    @staticmethod
+    def _normalize_inputs(
+        input: Union[AllEmbeddingInputValues, List[List[str]]]
+    ) -> List[List[str]]:
+        """
+        The Voyage contextualized embeddings API expects ``inputs`` to be a
+        list of documents, where each document is a ``list[str]`` of chunks.
+
+        Callers commonly pass a flat ``list[str]`` (the standard OpenAI
+        embedding input). Wrap it into a single document so the API always
+        receives ``list[list[str]]`` with ``list[str]`` elements.
+        """
+        # A single string -> one document with one chunk.
+        if isinstance(input, str):
+            return [[input]]
+        if isinstance(input, list):
+            # Already list[list[str]] -> pass through unchanged.
+            if input and all(isinstance(item, list) for item in input):
+                return input  # type: ignore[return-value]
+            # Flat list[str] -> wrap as a single document of chunks.
+            if all(isinstance(item, str) for item in input):
+                return [list(input)]  # type: ignore[arg-type]
+        return input  # type: ignore[return-value]
+
     def transform_embedding_response(
         self,
         model: str,

litellm/model_prices_and_context_window_backup.json

@@ -27345,6 +27345,38 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-4": {
+        "input_cost_per_token": 6e-08,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-large": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-lite": {
+        "input_cost_per_token": 2e-08,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-nano": {
+        "input_cost_per_token": 0.0,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "voyage/voyage-code-2": {
         "input_cost_per_token": 1.2e-07,
         "litellm_provider": "voyage",
@@ -27369,6 +27401,14 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-context-4": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 120000,
+        "max_tokens": 120000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "voyage/voyage-finance-2": {
         "input_cost_per_token": 1.2e-07,
         "litellm_provider": "voyage",
@@ -27417,6 +27457,14 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-multimodal-3.5": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "wandb/openai/gpt-oss-120b": {
         "max_tokens": 131072,
         "max_input_tokens": 131072,

model_prices_and_context_window.json

@@ -27345,6 +27345,38 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-4": {
+        "input_cost_per_token": 6e-08,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-large": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-lite": {
+        "input_cost_per_token": 2e-08,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
+    "voyage/voyage-4-nano": {
+        "input_cost_per_token": 0.0,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "voyage/voyage-code-2": {
         "input_cost_per_token": 1.2e-07,
         "litellm_provider": "voyage",
@@ -27369,6 +27401,14 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-context-4": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 120000,
+        "max_tokens": 120000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "voyage/voyage-finance-2": {
         "input_cost_per_token": 1.2e-07,
         "litellm_provider": "voyage",
@@ -27417,6 +27457,14 @@
         "mode": "embedding",
         "output_cost_per_token": 0.0
     },
+    "voyage/voyage-multimodal-3.5": {
+        "input_cost_per_token": 1.2e-07,
+        "litellm_provider": "voyage",
+        "max_input_tokens": 32000,
+        "max_tokens": 32000,
+        "mode": "embedding",
+        "output_cost_per_token": 0.0
+    },
     "wandb/openai/gpt-oss-120b": {
         "max_tokens": 131072,
         "max_input_tokens": 131072,

tests/llm_translation/test_voyage_ai.py

@@ -431,3 +431,28 @@ def test_contextual_embedding_multiple_inputs(self):
 
         except Exception as e:
             pytest.fail(f"Error occurred: {e}")
+
+
+class TestVoyageModelPrices:
+    """Ensure newly added Voyage models are registered with correct pricing"""
+
+    @pytest.mark.parametrize(
+        "model, input_cost, max_tokens",
+        [
+            ("voyage/voyage-4", 6e-08, 32000),
+            ("voyage/voyage-4-large", 1.2e-07, 32000),
+            ("voyage/voyage-4-lite", 2e-08, 32000),
+            ("voyage/voyage-4-nano", 0.0, 32000),
+            ("voyage/voyage-context-4", 1.2e-07, 120000),
+            ("voyage/voyage-multimodal-3.5", 1.2e-07, 32000),
+        ],
+    )
+    def test_voyage_model_registered(self, model, input_cost, max_tokens):
+        os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
+        litellm.model_cost = litellm.get_model_cost_map(url="")
+
+        info = litellm.get_model_info(model)
+        assert info["litellm_provider"] == "voyage"
+        assert info["mode"] == "embedding"
+        assert info["input_cost_per_token"] == input_cost
+        assert info["max_input_tokens"] == max_tokens

tests/test_litellm/llms/voyage/embedding/test_voyage_contextual_transformation.py

@@ -0,0 +1,60 @@
+import os
+import sys
+
+sys.path.insert(0, os.path.abspath("../../../../.."))
+
+from litellm.llms.voyage.embedding.transformation_contextual import (
+    VoyageContextualEmbeddingConfig,
+)
+
+
+class TestVoyageContextualDetection:
+    def test_voyage_context_4_detected(self):
+        config = VoyageContextualEmbeddingConfig()
+        assert config.is_contextualized_embeddings("voyage-context-4") is True
+        assert config.is_contextualized_embeddings("voyage/voyage-context-4") is True
+
+    def test_non_contextual_models_not_detected(self):
+        config = VoyageContextualEmbeddingConfig()
+        assert config.is_contextualized_embeddings("voyage-4") is False
+        assert config.is_contextualized_embeddings("voyage-4-nano") is False
+        assert config.is_contextualized_embeddings("voyage-3.5") is False
+
+
+class TestVoyageContextualInputNormalization:
+    """
+    The contextualized embeddings API expects ``inputs`` to be
+    ``list[list[str]]`` (each document is a ``list[str]`` of chunks).
+    """
+
+    def test_flat_list_of_str_is_wrapped(self):
+        config = VoyageContextualEmbeddingConfig()
+        transformed = config.transform_embedding_request(
+            "voyage-context-4", ["hello", "world"], {}, {}
+        )
+        assert transformed["inputs"] == [["hello", "world"]]
+        assert all(isinstance(doc, list) for doc in transformed["inputs"])
+
+    def test_single_string_is_wrapped(self):
+        config = VoyageContextualEmbeddingConfig()
+        transformed = config.transform_embedding_request(
+            "voyage-context-4", "hello", {}, {}
+        )
+        assert transformed["inputs"] == [["hello"]]
+
+    def test_nested_list_passed_through(self):
+        config = VoyageContextualEmbeddingConfig()
+        nested = [["a", "b"], ["c"]]
+        transformed = config.transform_embedding_request(
+            "voyage-context-4", nested, {}, {}
+        )
+        assert transformed["inputs"] == nested
+
+    def test_model_and_optional_params_preserved(self):
+        config = VoyageContextualEmbeddingConfig()
+        transformed = config.transform_embedding_request(
+            "voyage-context-4", ["x"], {"output_dimension": 512}, {}
+        )
+        assert transformed["model"] == "voyage-context-4"
+        assert transformed["output_dimension"] == 512
+        assert transformed["inputs"] == [["x"]]