diff --git a/nemo_retriever/helm/README.md b/nemo_retriever/helm/README.md
index 8059ffc3da..5ba9b9291c 100644
--- a/nemo_retriever/helm/README.md
+++ b/nemo_retriever/helm/README.md
@@ -331,6 +331,7 @@ The retriever service picks up the in-cluster ASR endpoint when `nimOperator.aud
 | `serviceConfig.vectordb.enabled`                  | `true`  | Deploy the LanceDB vectordb Pod. When `true` the chart **requires** a resolvable embed endpoint (refer to [VectorDB and the embed endpoint](#vectordb-and-the-embed-endpoint)); `helm install` / `helm upgrade` fails fast otherwise. |
 | `serviceConfig.vectordb.lancedbUri`               | `/data/vectordb` | LanceDB on the vectordb Pod's PVC. |
 | `serviceConfig.vectordb.embedModel`               | `nvidia/llama-nemotron-embed-vl-1b-v2` | Passed to vectordb + worker `embed_model_name`. |
+| `serviceConfig.vectordb.embedModelProviderPrefix` | `""` | Optional LiteLLM provider prefix prepended to the remote embed model name. |
 
 #### VectorDB and the embed endpoint { #vectordb-and-the-embed-endpoint }
 
diff --git a/nemo_retriever/helm/templates/configmap.yaml b/nemo_retriever/helm/templates/configmap.yaml
index a5c58bd773..ba04101745 100644
--- a/nemo_retriever/helm/templates/configmap.yaml
+++ b/nemo_retriever/helm/templates/configmap.yaml
@@ -67,6 +67,7 @@ nim_endpoints:
   ocr_invoke_url: {{ .ocrURL | quote }}
   embed_invoke_url: {{ .embedURL | quote }}
   embed_model_name: {{ .Values.serviceConfig.vectordb.embedModel | quote }}
+  embed_model_provider_prefix: {{ if .Values.serviceConfig.vectordb.embedModelProviderPrefix }}{{ .Values.serviceConfig.vectordb.embedModelProviderPrefix | quote }}{{ else }}null{{ end }}
   caption_invoke_url: {{ if .captionURL }}{{ .captionURL | quote }}{{ else }}null{{ end }}
   caption_model_name: {{ if .captionModelName }}{{ .captionModelName | quote }}{{ else }}null{{ end }}
   audio_grpc_endpoint: {{ if .audioGrpcEndpoint }}{{ .audioGrpcEndpoint | quote }}{{ else }}null{{ end }}
@@ -140,6 +141,7 @@ vectordb:
   lancedb_uri: {{ .Values.serviceConfig.vectordb.lancedbUri | quote }}
   table_name: {{ .Values.serviceConfig.vectordb.tableName | quote }}
   embed_model: {{ .Values.serviceConfig.vectordb.embedModel | quote }}
+  embed_model_provider_prefix: {{ if .Values.serviceConfig.vectordb.embedModelProviderPrefix }}{{ .Values.serviceConfig.vectordb.embedModelProviderPrefix | quote }}{{ else }}null{{ end }}
   vectordb_url: "http://{{ .vectordbSvc }}:{{ .vectordbPort }}"
 {{- else }}
 vectordb:
diff --git a/nemo_retriever/helm/templates/deployment-vectordb.yaml b/nemo_retriever/helm/templates/deployment-vectordb.yaml
index ebee3b2729..19c4efc009 100644
--- a/nemo_retriever/helm/templates/deployment-vectordb.yaml
+++ b/nemo_retriever/helm/templates/deployment-vectordb.yaml
@@ -95,6 +95,10 @@ spec:
             {{- end }}
             - --embed-model
             - {{ .Values.serviceConfig.vectordb.embedModel | default $localModels.embed.modelName | quote }}
+            {{- if .Values.serviceConfig.vectordb.embedModelProviderPrefix }}
+            - --embed-model-provider-prefix
+            - {{ .Values.serviceConfig.vectordb.embedModelProviderPrefix | quote }}
+            {{- end }}
             - --port
             - {{ $vdb.port | quote }}
             {{- if $embedURL }}
diff --git a/nemo_retriever/helm/values.yaml b/nemo_retriever/helm/values.yaml
index 7811bd20fe..685d7bfd2d 100644
--- a/nemo_retriever/helm/values.yaml
+++ b/nemo_retriever/helm/values.yaml
@@ -660,6 +660,7 @@ serviceConfig:
     lancedbUri: "/data/vectordb"
     tableName: "nemo_retriever"
     embedModel: "nvidia/llama-nemotron-embed-vl-1b-v2"
+    embedModelProviderPrefix: ""
 
   # Optional bearer-token authentication. When apiToken is set, every
   # request must carry "Authorization: Bearer <token>".
diff --git a/nemo_retriever/src/nemo_retriever/cli/ingest/graph_commands.py b/nemo_retriever/src/nemo_retriever/cli/ingest/graph_commands.py
index b0e36bedcb..2c5eac1b69 100644
--- a/nemo_retriever/src/nemo_retriever/cli/ingest/graph_commands.py
+++ b/nemo_retriever/src/nemo_retriever/cli/ingest/graph_commands.py
@@ -266,6 +266,7 @@ def _graph_ingest_command(
     table_output_format: opts.TableOutputFormatOption = None,
     embed_invoke_url: opts.EmbedInvokeUrlOption = None,
     embed_model_name: opts.EmbedModelNameOption = None,
+    embed_model_provider_prefix: opts.EmbedModelProviderPrefixOption = None,
     local_ingest_embed_backend: opts.LocalIngestEmbedBackendOption = None,
     embed_modality: opts.EmbedModalityOption = None,
     embed_granularity: opts.EmbedGranularityOption = None,
diff --git a/nemo_retriever/src/nemo_retriever/cli/ingest/options.py b/nemo_retriever/src/nemo_retriever/cli/ingest/options.py
index 4ccd0c5cfc..4115d3e555 100644
--- a/nemo_retriever/src/nemo_retriever/cli/ingest/options.py
+++ b/nemo_retriever/src/nemo_retriever/cli/ingest/options.py
@@ -288,6 +288,13 @@
         help=f"Optional embedding model name override. Defaults to {DEFAULT_EMBED_MODEL} when omitted.",
     ),
 ]
+EmbedModelProviderPrefixOption = Annotated[
+    str | None,
+    typer.Option(
+        "--embed-model-provider-prefix",
+        help="Optional LiteLLM provider prefix prepended to the remote embedding model name.",
+    ),
+]
 LocalIngestEmbedBackendOption = Annotated[
     LocalIngestEmbedBackendValue | None,
     typer.Option(
diff --git a/nemo_retriever/src/nemo_retriever/cli/query/app.py b/nemo_retriever/src/nemo_retriever/cli/query/app.py
index 404c8c14d8..9ae7455537 100644
--- a/nemo_retriever/src/nemo_retriever/cli/query/app.py
+++ b/nemo_retriever/src/nemo_retriever/cli/query/app.py
@@ -174,6 +174,7 @@ def _local_command(
     table_name: opts.TableNameOption = "nemo-retriever",
     embed_invoke_url: opts.EmbedInvokeUrlOption = None,
     embed_model_name: opts.EmbedModelNameOption = None,
+    embed_model_provider_prefix: opts.EmbedModelProviderPrefixOption = None,
     reranker_invoke_url: opts.RerankerInvokeUrlOption = None,
     reranker_api_key_env: opts.RerankerApiKeyEnvOption = None,
     reranker_model_name: opts.RerankerModelNameOption = None,
@@ -220,6 +221,7 @@ def _local_command(
                 embed=QueryEmbedOptions(
                     embed_invoke_url=embed_invoke_url,
                     embed_model_name=embed_model_name,
+                    embed_model_provider_prefix=embed_model_provider_prefix,
                 ),
                 rerank=QueryRerankOptions(
                     enabled=rerank,
@@ -261,6 +263,7 @@ def _request() -> QueryRequest:
                 embed=QueryEmbedOptions(
                     embed_invoke_url=embed_invoke_url,
                     embed_model_name=embed_model_name,
+                    embed_model_provider_prefix=embed_model_provider_prefix,
                 ),
                 rerank=QueryRerankOptions(
                     enabled=rerank,
diff --git a/nemo_retriever/src/nemo_retriever/cli/query/options.py b/nemo_retriever/src/nemo_retriever/cli/query/options.py
index 0c60b78d2e..089e7e26dc 100644
--- a/nemo_retriever/src/nemo_retriever/cli/query/options.py
+++ b/nemo_retriever/src/nemo_retriever/cli/query/options.py
@@ -72,6 +72,13 @@
         help=f"Optional embedding model name override. Defaults to {DEFAULT_EMBED_MODEL} when omitted.",
     ),
 ]
+EmbedModelProviderPrefixOption = Annotated[
+    str | None,
+    typer.Option(
+        "--embed-model-provider-prefix",
+        help="Optional LiteLLM provider prefix prepended to the remote embedding model name.",
+    ),
+]
 RerankerInvokeUrlOption = Annotated[
     str | None,
     typer.Option("--reranker-invoke-url", help="Reranker endpoint URL."),
diff --git a/nemo_retriever/src/nemo_retriever/common/api/util/string_processing/__init__.py b/nemo_retriever/src/nemo_retriever/common/api/util/string_processing/__init__.py
index 6fb7fb5e45..7182944dba 100644
--- a/nemo_retriever/src/nemo_retriever/common/api/util/string_processing/__init__.py
+++ b/nemo_retriever/src/nemo_retriever/common/api/util/string_processing/__init__.py
@@ -51,6 +51,17 @@ def ensure_openai_embeddings_http_url(endpoint_url: str) -> str:
     return urlunsplit((parts.scheme, parts.netloc, new_path, parts.query, parts.fragment))
 
 
+def prepend_model_provider_prefix(model_name: str | None, model_provider_prefix: str | None) -> str | None:
+    """Prepend a LiteLLM provider prefix to a model identifier when configured."""
+    if model_name is None:
+        return None
+    model = str(model_name).strip()
+    prefix = str(model_provider_prefix or "").strip().strip("/")
+    if not model or not prefix:
+        return model
+    return f"{prefix}/{model.lstrip('/')}"
+
+
 def generate_url(url) -> str:
     """Examines the user defined URL for http*://. If that
     pattern is detected the URL is used as provided by the user.
diff --git a/nemo_retriever/src/nemo_retriever/common/params/models.py b/nemo_retriever/src/nemo_retriever/common/params/models.py
index 71c046cee8..3112249bfb 100644
--- a/nemo_retriever/src/nemo_retriever/common/params/models.py
+++ b/nemo_retriever/src/nemo_retriever/common/params/models.py
@@ -384,6 +384,7 @@ class EmbedParams(_ParamsModel):
     embedding_endpoint: Optional[str] = None
     embed_invoke_url: Optional[str] = None
     embed_model_name: Optional[str] = None
+    embed_model_provider_prefix: Optional[str] = None
     api_key: Optional[str] = None
     input_type: str = "passage"
     embed_modality: str = "text"  # "text", "image", or "text_image" — default for all element types
diff --git a/nemo_retriever/src/nemo_retriever/common/params/utils.py b/nemo_retriever/src/nemo_retriever/common/params/utils.py
index f7c82f7c88..bf1b9b2d36 100644
--- a/nemo_retriever/src/nemo_retriever/common/params/utils.py
+++ b/nemo_retriever/src/nemo_retriever/common/params/utils.py
@@ -8,6 +8,8 @@
 
 from typing import TYPE_CHECKING, Any, Dict
 
+from nemo_retriever.common.api.util.string_processing import prepend_model_provider_prefix
+
 if TYPE_CHECKING:
     from nemo_retriever.common.params.models import BatchTuningParams
 
@@ -50,6 +52,12 @@ def normalize_embed_kwargs(kwargs: Dict[str, Any]) -> Dict[str, Any]:
 
     if "embed_invoke_url" in normalized:
         normalized.setdefault("embedding_endpoint", normalized["embed_invoke_url"])
+    endpoint = normalized.get("embedding_endpoint") or normalized.get("embed_invoke_url")
+    model_provider_prefix = normalized.pop("embed_model_provider_prefix", None)
+    if endpoint and model_provider_prefix:
+        for key in ("model_name", "embed_model_name"):
+            if key in normalized:
+                normalized[key] = prepend_model_provider_prefix(normalized[key], str(model_provider_prefix))
     return normalized
 
 
@@ -58,6 +66,7 @@ def build_embed_option_kwargs(
     embed_model_name: str | None,
     local_ingest_embed_backend: str | None = None,
     embed_api_key: str | None = None,
+    embed_model_provider_prefix: str | None = None,
     embed_modality: str | None = None,
     text_elements_modality: str | None = None,
     structured_elements_modality: str | None = None,
@@ -79,6 +88,8 @@ def build_embed_option_kwargs(
         embed_kwargs["local_ingest_embed_backend"] = local_ingest_embed_backend
     if embed_api_key is not None:
         embed_kwargs["api_key"] = embed_api_key
+    if embed_model_provider_prefix is not None:
+        embed_kwargs["embed_model_provider_prefix"] = embed_model_provider_prefix
     if embed_modality is not None:
         embed_kwargs["embed_modality"] = embed_modality
     if text_elements_modality is not None:
diff --git a/nemo_retriever/src/nemo_retriever/ingest/plan.py b/nemo_retriever/src/nemo_retriever/ingest/plan.py
index e857f4385f..f6ff01f6e7 100644
--- a/nemo_retriever/src/nemo_retriever/ingest/plan.py
+++ b/nemo_retriever/src/nemo_retriever/ingest/plan.py
@@ -185,6 +185,7 @@ class IngestEmbedBatchOptions:
 class IngestEmbedOptions:
     embed_invoke_url: str | None = None
     embed_model_name: str | None = None
+    embed_model_provider_prefix: str | None = None
     local_ingest_embed_backend: LocalIngestEmbedBackendValue | None = None
     embed_api_key: str | None = None
     embed_modality: str | None = None
@@ -642,6 +643,7 @@ def resolve_ingest_plan(request: IngestPlanRequest) -> ResolvedIngestPlan:
         embed.embed_model_name,
         local_ingest_embed_backend=embed.local_ingest_embed_backend,
         embed_api_key=embed.embed_api_key,
+        embed_model_provider_prefix=embed.embed_model_provider_prefix,
         embed_modality=embed.embed_modality,
         text_elements_modality=embed.text_elements_modality,
         structured_elements_modality=embed.structured_elements_modality,
diff --git a/nemo_retriever/src/nemo_retriever/models/inference/main_text_embed.py b/nemo_retriever/src/nemo_retriever/models/inference/main_text_embed.py
index dd7c29c373..99fa2cde0a 100644
--- a/nemo_retriever/src/nemo_retriever/models/inference/main_text_embed.py
+++ b/nemo_retriever/src/nemo_retriever/models/inference/main_text_embed.py
@@ -48,7 +48,10 @@ def local_embedder(texts):
 from typing import Any, Callable, Dict, Iterable, List, Optional, Sequence, Tuple
 
 import pandas as pd
-from nemo_retriever.common.api.util.string_processing import ensure_openai_embeddings_http_url
+from nemo_retriever.common.api.util.string_processing import (
+    ensure_openai_embeddings_http_url,
+    prepend_model_provider_prefix,
+)
 
 from nemo_retriever.models import _DEFAULT_EMBED_MODEL
 from nemo_retriever.common.params.models import IMAGE_MODALITIES
@@ -72,6 +75,7 @@ class TextEmbeddingConfig:
     api_key: Optional[str] = None
     embedding_nim_endpoint: Optional[str] = None  # e.g. "http://host:8000/v1"
     embedding_model: str = _DEFAULT_EMBED_MODEL
+    embedding_model_provider_prefix: Optional[str] = None
     encoding_format: str = "float"  # OpenAI-compatible embeddings often accept "float"
     input_type: str = "passage"
     truncate: str = "END"
@@ -293,6 +297,7 @@ def _http_embed_openai_compat(
     encoding_format: str,
     input_type: str,
     truncate: str,
+    model_provider_prefix: Optional[str] = None,
     dimensions: Optional[int] = None,
     timeout_s: float = 600.0,
 ) -> List[Optional[List[float]]]:
@@ -308,6 +313,7 @@ def _http_embed_openai_compat(
         raise RuntimeError("Remote embedding requested but `httpx` is not installed.") from e
 
     url = _normalize_embeddings_endpoint(_pick_embed_endpoint(endpoint_url))
+    model_name = prepend_model_provider_prefix(model_name, model_provider_prefix) or model_name
     headers: Dict[str, str] = {"accept": "application/json", "content-type": "application/json"}
     token = (api_key or "").strip()
     if token:
@@ -353,6 +359,7 @@ def _make_async_request(
     api_key: Optional[str],
     embedding_nim_endpoint: str,
     embedding_model: str,
+    embedding_model_provider_prefix: Optional[str],
     encoding_format: str,
     input_type: str,
     truncate: str,
@@ -377,6 +384,7 @@ def _make_async_request(
             api_key=api_key,
             endpoint_url=str(embedding_nim_endpoint),
             model_name=str(embedding_model),
+            model_provider_prefix=embedding_model_provider_prefix,
             encoding_format=str(encoding_format),
             input_type=str(input_type),
             truncate=str(truncate),
@@ -399,6 +407,7 @@ def _async_request_handler(
     api_key: Optional[str],
     embedding_nim_endpoint: str,
     embedding_model: str,
+    embedding_model_provider_prefix: Optional[str],
     encoding_format: str,
     input_type: str,
     truncate: str,
@@ -420,6 +429,7 @@ def _async_request_handler(
                 api_key=api_key or None,
                 embedding_nim_endpoint=str(embedding_nim_endpoint),
                 embedding_model=str(embedding_model),
+                embedding_model_provider_prefix=embedding_model_provider_prefix,
                 encoding_format=str(encoding_format),
                 input_type=str(input_type),
                 truncate=str(truncate),
@@ -440,6 +450,7 @@ def _async_runner(
     api_key: Optional[str],
     embedding_nim_endpoint: str,
     embedding_model: str,
+    embedding_model_provider_prefix: Optional[str],
     encoding_format: str,
     input_type: str,
     truncate: str,
@@ -454,6 +465,7 @@ def _async_runner(
         api_key,
         embedding_nim_endpoint,
         embedding_model,
+        embedding_model_provider_prefix,
         encoding_format,
         input_type,
         truncate,
@@ -580,6 +592,11 @@ def create_text_embeddings_for_df(
         task_config["endpoint_url"] if "endpoint_url" in task_config else transform_config.embedding_nim_endpoint
     )
     model_name = task_config["model_name"] if "model_name" in task_config else transform_config.embedding_model
+    model_provider_prefix = (
+        task_config["model_provider_prefix"]
+        if "model_provider_prefix" in task_config
+        else task_config.get("embed_model_provider_prefix", transform_config.embedding_model_provider_prefix)
+    )
     dimensions = task_config["dimensions"] if "dimensions" in task_config else transform_config.dimensions
 
     endpoint_url = endpoint_url.strip() if isinstance(endpoint_url, str) else endpoint_url
@@ -671,6 +688,7 @@ def _text_image_content(r: pd.Series) -> Optional[str]:
                 api_key,
                 str(endpoint_url),
                 str(model_name),
+                model_provider_prefix,
                 str(transform_config.encoding_format),
                 str(transform_config.input_type),
                 str(transform_config.truncate),
@@ -693,6 +711,7 @@ def _text_image_content(r: pd.Series) -> Optional[str]:
                     api_key,
                     str(endpoint_url),
                     str(model_name),
+                    model_provider_prefix,
                     str(transform_config.encoding_format),
                     str(transform_config.input_type),
                     str(transform_config.truncate),
diff --git a/nemo_retriever/src/nemo_retriever/models/inference/runtime.py b/nemo_retriever/src/nemo_retriever/models/inference/runtime.py
index 1206cbcee2..2edd8b9cd4 100644
--- a/nemo_retriever/src/nemo_retriever/models/inference/runtime.py
+++ b/nemo_retriever/src/nemo_retriever/models/inference/runtime.py
@@ -30,6 +30,7 @@ def _embed_group(
     inference_batch_size: int,
     output_column: str,
     resolved_model_name: str,
+    embed_model_provider_prefix: Optional[str] = None,
     nim_http_max_concurrent: int = 32,
     input_type: str = "passage",
     request_timeout_s: float = 600.0,
@@ -76,6 +77,7 @@ def embedder(texts: Sequence[str]) -> Sequence[Sequence[float]]:  # noqa
         dimensions=None,
         embedding_nim_endpoint=endpoint or "http://localhost:8012/v1",
         embedding_model=resolved_model_name or VL_EMBED_MODEL,
+        embedding_model_provider_prefix=embed_model_provider_prefix,
         embed_modality=group_modality,
         nim_http_max_concurrent=max(1, int(nim_http_max_concurrent)),
     )
@@ -87,6 +89,7 @@ def embedder(texts: Sequence[str]) -> Sequence[Sequence[float]]:  # noqa
             "embedder": embedder,
             "multimodal_embedder": multimodal_embedder,
             "endpoint_url": endpoint,
+            "embed_model_provider_prefix": embed_model_provider_prefix,
             "local_batch_size": int(effective_batch_size),
             "nim_http_max_concurrent": max(1, int(nim_http_max_concurrent)),
             "request_timeout_s": float(request_timeout_s),
@@ -110,6 +113,7 @@ def embed_text_main_text_embed(
     embedding_dim_column: str = "text_embeddings_1b_v2_dim",
     has_embedding_column: str = "text_embeddings_1b_v2_has_embedding",
     embed_modality: str = "text",
+    embed_model_provider_prefix: Optional[str] = None,
     nim_http_max_concurrent: int = 32,
     input_type: str = "passage",
     request_timeout_s: float | None = None,
@@ -148,6 +152,7 @@ def embed_text_main_text_embed(
                 inference_batch_size=inference_batch_size,
                 output_column=output_column,
                 resolved_model_name=resolved_model_name,
+                embed_model_provider_prefix=embed_model_provider_prefix,
                 nim_http_max_concurrent=nim_http_max_concurrent,
                 input_type=input_type,
                 request_timeout_s=float(request_timeout_s),
@@ -169,6 +174,7 @@ def embed_text_main_text_embed(
                     inference_batch_size=inference_batch_size,
                     output_column=output_column,
                     resolved_model_name=resolved_model_name,
+                    embed_model_provider_prefix=embed_model_provider_prefix,
                     nim_http_max_concurrent=nim_http_max_concurrent,
                     input_type=input_type,
                     request_timeout_s=float(request_timeout_s),
diff --git a/nemo_retriever/src/nemo_retriever/models/inference/shared.py b/nemo_retriever/src/nemo_retriever/models/inference/shared.py
index 21d042dbfd..22066a1c01 100644
--- a/nemo_retriever/src/nemo_retriever/models/inference/shared.py
+++ b/nemo_retriever/src/nemo_retriever/models/inference/shared.py
@@ -7,6 +7,7 @@
 from __future__ import annotations
 
 from nemo_retriever.common.params import EmbedParams
+from nemo_retriever.common.params.utils import normalize_embed_kwargs
 
 
 def _to_bool(v: object, default: bool = False) -> bool:
@@ -22,6 +23,4 @@ def build_embed_kwargs(params: EmbedParams) -> dict[str, object]:
         **params.model_dump(mode="python", exclude={"runtime", "batch_tuning"}, exclude_none=True),
         **params.runtime.model_dump(mode="python", exclude_none=True),
     }
-    if "embedding_endpoint" not in kwargs and kwargs.get("embed_invoke_url"):
-        kwargs["embedding_endpoint"] = kwargs.get("embed_invoke_url")
-    return kwargs
+    return normalize_embed_kwargs(kwargs)
diff --git a/nemo_retriever/src/nemo_retriever/models/nim/util/__init__.py b/nemo_retriever/src/nemo_retriever/models/nim/util/__init__.py
index 20394de67f..e07ddcad36 100644
--- a/nemo_retriever/src/nemo_retriever/models/nim/util/__init__.py
+++ b/nemo_retriever/src/nemo_retriever/models/nim/util/__init__.py
@@ -13,7 +13,10 @@
 from nemo_retriever.models.nim.primitives.model_interface.text_embedding import EmbeddingModelInterface
 from nemo_retriever.models.nim.primitives.nim_client import get_nim_client_manager
 from nemo_retriever.models.nim.primitives.nim_model_interface import ModelInterface
-from nemo_retriever.common.api.util.string_processing import ensure_openai_embeddings_http_url
+from nemo_retriever.common.api.util.string_processing import (
+    ensure_openai_embeddings_http_url,
+    prepend_model_provider_prefix,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -138,6 +141,7 @@ def infer_microservice(
     input_type: str = "passage",
     truncate: str = "END",
     batch_size: int = 8191,
+    model_provider_prefix: str | None = None,
     grpc: bool = False,
     input_names: list = ["text"],
     output_names: list = ["embeddings"],
@@ -200,6 +204,7 @@ def infer_microservice(
         )
     else:
         embedding_endpoint = ensure_openai_embeddings_http_url(str(embedding_endpoint))
+        model_name = prepend_model_provider_prefix(model_name, model_provider_prefix)
         client = NimClient(
             model_interface=EmbeddingModelInterface(),
             protocol="http",
diff --git a/nemo_retriever/src/nemo_retriever/query/agentic.py b/nemo_retriever/src/nemo_retriever/query/agentic.py
index aa75034470..b0d3fe5adc 100644
--- a/nemo_retriever/src/nemo_retriever/query/agentic.py
+++ b/nemo_retriever/src/nemo_retriever/query/agentic.py
@@ -113,6 +113,7 @@ class AgenticRetrievalConfig:
     vdb_op: str = "lancedb"
     vdb_kwargs: dict[str, Any] = field(default_factory=dict)
     query_embedder: str = VL_EMBED_MODEL
+    query_embedder_provider_prefix: Optional[str] = None
     embedding_endpoint: Optional[str] = None
     embedding_api_key: str = ""
     local_hf_batch_size: int = 32
@@ -174,6 +175,7 @@ def __init__(
             embed_kwargs={
                 "model_name": str(cfg.query_embedder or VL_EMBED_MODEL),
                 "embed_model_name": str(cfg.query_embedder or VL_EMBED_MODEL),
+                "embed_model_provider_prefix": cfg.query_embedder_provider_prefix,
                 "embedding_endpoint": cfg.embedding_endpoint,
                 "api_key": cfg.embedding_api_key,
                 "input_type": "query",
diff --git a/nemo_retriever/src/nemo_retriever/query/options.py b/nemo_retriever/src/nemo_retriever/query/options.py
index 237001a6e9..abb978fddd 100644
--- a/nemo_retriever/src/nemo_retriever/query/options.py
+++ b/nemo_retriever/src/nemo_retriever/query/options.py
@@ -26,6 +26,7 @@ class QueryRetrievalOptions:
 class QueryEmbedOptions:
     embed_invoke_url: str | None = None
     embed_model_name: str | None = None
+    embed_model_provider_prefix: str | None = None
 
 
 @dataclass(frozen=True)
diff --git a/nemo_retriever/src/nemo_retriever/query/workflow.py b/nemo_retriever/src/nemo_retriever/query/workflow.py
index 5c777d5378..c8af172819 100644
--- a/nemo_retriever/src/nemo_retriever/query/workflow.py
+++ b/nemo_retriever/src/nemo_retriever/query/workflow.py
@@ -102,7 +102,11 @@ def query_kwargs(self) -> dict[str, Any]:
 
 def resolve_query_plan(request: QueryRequest) -> ResolvedQueryPlan:
     """Resolve root query options once so callers can reuse a Retriever."""
-    embed_kwargs = build_embed_option_kwargs(request.embed.embed_invoke_url, request.embed.embed_model_name)
+    embed_kwargs = build_embed_option_kwargs(
+        request.embed.embed_invoke_url,
+        request.embed.embed_model_name,
+        embed_model_provider_prefix=request.embed.embed_model_provider_prefix,
+    )
     rerank_kwargs = _build_rerank_kwargs(request.rerank) if request.rerank.enabled else {}
     content_types = request.retrieval.content_types
     if content_types is not None and not isinstance(content_types, str):
@@ -179,6 +183,8 @@ def agentic_query_documents(request: QueryRequest) -> list[dict[str, Any]]:
     }
     if request.embed.embed_model_name:
         cfg_kwargs["query_embedder"] = request.embed.embed_model_name
+    if request.embed.embed_model_provider_prefix:
+        cfg_kwargs["query_embedder_provider_prefix"] = request.embed.embed_model_provider_prefix
     if request.rerank.enabled:
         # `reranker` doubles as the on/off gate (rerank=bool(cfg.reranker)) and the
         # model name, so fall back to the default model when only --rerank is given.
diff --git a/nemo_retriever/src/nemo_retriever/service/config.py b/nemo_retriever/src/nemo_retriever/service/config.py
index 08f306608b..e7e3010792 100644
--- a/nemo_retriever/src/nemo_retriever/service/config.py
+++ b/nemo_retriever/src/nemo_retriever/service/config.py
@@ -141,6 +141,13 @@ class NimEndpointsConfig(RichModel):
             "Server-owned — clients cannot override the deployed embed NIM SKU."
         ),
     )
+    embed_model_provider_prefix: str | None = Field(
+        default=None,
+        description=(
+            "Optional LiteLLM provider prefix prepended to embed_model_name for "
+            "remote embedding endpoints that require namespaced model IDs."
+        ),
+    )
     rerank_invoke_url: str | None = None
     audio_grpc_endpoint: str | None = Field(
         default=None,
@@ -293,6 +300,7 @@ class VectorDbConfig(RichModel):
     lancedb_uri: str = "/data/vectordb"
     table_name: str = "nemo_retriever"
     embed_model: str = "nvidia/llama-nemotron-embed-vl-1b-v2"
+    embed_model_provider_prefix: str | None = None
     vectordb_url: str = Field(
         default="http://nemo-retriever-vectordb:7671",
         description="URL of the vectordb service (for workers to POST embeddings to)",
diff --git a/nemo_retriever/src/nemo_retriever/service/retriever-service.yaml b/nemo_retriever/src/nemo_retriever/service/retriever-service.yaml
index 3a797bf7db..449f53f1fa 100644
--- a/nemo_retriever/src/nemo_retriever/service/retriever-service.yaml
+++ b/nemo_retriever/src/nemo_retriever/service/retriever-service.yaml
@@ -37,6 +37,9 @@ nim_endpoints:
   embed_invoke_url: null
   # Model name for the remote embed NIM (server-owned; must match the SKU).
   embed_model_name: null
+  # Optional LiteLLM provider prefix prepended to embed_model_name for
+  # proxies that require provider/model IDs.
+  embed_model_provider_prefix: null
   # gRPC endpoint for the Parakeet ASR NIM (e.g. parakeet-nim:50051).
   # When set, audio/video pipelines use remote ASR instead of loading
   # the local Parakeet model (which requires torch + GPU).
diff --git a/nemo_retriever/src/nemo_retriever/service/service_ingestor.py b/nemo_retriever/src/nemo_retriever/service/service_ingestor.py
index e53112c641..af4d5cbb47 100644
--- a/nemo_retriever/src/nemo_retriever/service/service_ingestor.py
+++ b/nemo_retriever/src/nemo_retriever/service/service_ingestor.py
@@ -219,6 +219,7 @@ def _normalize_files(files: Union[str, List[str], List[Path]]) -> list[Path]:
         "nemotron_parse_invoke_url",
         "embed_invoke_url",
         "embedding_endpoint",
+        "embed_model_provider_prefix",
         "endpoint_url",
         "api_base",
         "auth_token",
diff --git a/nemo_retriever/src/nemo_retriever/service/services/pipeline_executor.py b/nemo_retriever/src/nemo_retriever/service/services/pipeline_executor.py
index 2e01fae0f4..e9f8c53b81 100644
--- a/nemo_retriever/src/nemo_retriever/service/services/pipeline_executor.py
+++ b/nemo_retriever/src/nemo_retriever/service/services/pipeline_executor.py
@@ -318,6 +318,7 @@ def _post_rows_to_vectordb(rows: list[dict[str, Any]], vectordb_url: str, filena
     "embedding_endpoint",
     "api_key",
     "embed_model_name",
+    "embed_model_provider_prefix",
     "model_name",
 )
 # Trust-owned caption keys. ``endpoint_url`` / ``api_key`` /
@@ -851,6 +852,8 @@ def build_embed_params(nim: "NimEndpointsConfig", local: "LocalModelsConfig | No
         if nim.embed_model_name:
             kwargs["model_name"] = nim.embed_model_name
             kwargs["embed_model_name"] = nim.embed_model_name
+        if nim.embed_model_provider_prefix:
+            kwargs["embed_model_provider_prefix"] = nim.embed_model_provider_prefix
         if nim.api_key:
             kwargs["api_key"] = nim.api_key
         return EmbedParams(**kwargs)
diff --git a/nemo_retriever/src/nemo_retriever/service/vectordb_app.py b/nemo_retriever/src/nemo_retriever/service/vectordb_app.py
index 8221d2c298..dd1ece41c5 100644
--- a/nemo_retriever/src/nemo_retriever/service/vectordb_app.py
+++ b/nemo_retriever/src/nemo_retriever/service/vectordb_app.py
@@ -90,12 +90,14 @@ def _embed_queries_remote(
     embed_model: str,
     embed_endpoint: str,
     embed_api_key: str,
+    embed_model_provider_prefix: str | None = None,
 ) -> list[list[float]]:
     from nemo_retriever.models.nim.util import infer_microservice
 
     return infer_microservice(
         texts,
         model_name=embed_model,
+        model_provider_prefix=embed_model_provider_prefix,
         embedding_endpoint=embed_endpoint,
         nvidia_api_key=embed_api_key or None,
         input_type="query",
@@ -117,6 +119,7 @@ def __init__(
         embed_model: str,
         embed_api_key: str,
         *,
+        embed_model_provider_prefix: str | None = None,
         local_embed: bool = False,
         local_embed_backend: str = "hf",
         hf_cache_dir: str | None = None,
@@ -127,6 +130,7 @@ def __init__(
         self.table_name = table_name
         self.embed_endpoint = embed_endpoint
         self.embed_model = embed_model
+        self.embed_model_provider_prefix = embed_model_provider_prefix
         self.embed_api_key = embed_api_key
         self.local_embed = local_embed
         self.local_embed_backend = local_embed_backend
@@ -244,6 +248,7 @@ def embed_queries(self, texts: list[str]) -> list[list[float]]:
             return _embed_queries_remote(
                 texts,
                 embed_model=self.embed_model,
+                embed_model_provider_prefix=self.embed_model_provider_prefix,
                 embed_endpoint=self.embed_endpoint,
                 embed_api_key=self.embed_api_key,
             )
@@ -268,6 +273,7 @@ def create_vectordb_app(
     table_name: str = "nemo_retriever",
     embed_endpoint: str = "",
     embed_model: str = "nvidia/llama-nemotron-embed-vl-1b-v2",
+    embed_model_provider_prefix: str | None = None,
     embed_api_key: str = "",
     *,
     local_embed: bool = False,
@@ -286,6 +292,7 @@ async def lifespan(app: FastAPI) -> AsyncIterator[None]:
             table_name=table_name,
             embed_endpoint=embed_endpoint,
             embed_model=embed_model,
+            embed_model_provider_prefix=embed_model_provider_prefix,
             embed_api_key=embed_api_key,
             local_embed=local_embed,
             local_embed_backend=local_embed_backend,
@@ -384,6 +391,7 @@ def main() -> None:
     parser.add_argument("--table-name", default="nemo_retriever", help="LanceDB table name")
     parser.add_argument("--embed-endpoint", default="", help="Remote NIM/OpenAI-compatible embed URL")
     parser.add_argument("--embed-model", default="nvidia/llama-nemotron-embed-vl-1b-v2")
+    parser.add_argument("--embed-model-provider-prefix", default="", help="Optional LiteLLM provider prefix")
     parser.add_argument("--embed-api-key", default="")
     parser.add_argument(
         "--local-embed",
@@ -422,6 +430,7 @@ def main() -> None:
         table_name=args.table_name,
         embed_endpoint=args.embed_endpoint,
         embed_model=args.embed_model,
+        embed_model_provider_prefix=args.embed_model_provider_prefix or None,
         embed_api_key=args.embed_api_key,
         local_embed=args.local_embed,
         local_embed_backend=args.local_embed_backend,
diff --git a/nemo_retriever/tests/test_embed_params.py b/nemo_retriever/tests/test_embed_params.py
index 596c1b06fc..db2b334c99 100644
--- a/nemo_retriever/tests/test_embed_params.py
+++ b/nemo_retriever/tests/test_embed_params.py
@@ -11,6 +11,7 @@
 import pytest
 
 from nemo_retriever.common.params.models import EmbedParams, IMAGE_MODALITIES
+from nemo_retriever.common.params.utils import build_embed_option_kwargs
 
 
 def test_image_text_alias_is_rejected():
@@ -54,6 +55,62 @@ def test_image_modalities_constant():
     assert isinstance(IMAGE_MODALITIES, frozenset)
 
 
+def test_build_embed_option_kwargs_applies_remote_model_provider_prefix():
+    kwargs = build_embed_option_kwargs(
+        "https://litellm.example.com/v1/embeddings",
+        "nvidia/llama-nemotron-embed-vl-1b-v2",
+        embed_model_provider_prefix="nvidia",
+    )
+
+    assert kwargs["model_name"] == "nvidia/nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert kwargs["embed_model_name"] == "nvidia/nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert "embed_model_provider_prefix" not in kwargs
+
+
+def test_build_embed_option_kwargs_leaves_model_unchanged_without_prefix():
+    kwargs = build_embed_option_kwargs(
+        "https://integrate.api.nvidia.com/v1/embeddings",
+        "nvidia/llama-nemotron-embed-vl-1b-v2",
+    )
+
+    assert kwargs["model_name"] == "nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert kwargs["embed_model_name"] == "nvidia/llama-nemotron-embed-vl-1b-v2"
+
+
+def test_build_embed_option_kwargs_prefix_supports_other_vendor_namespaces():
+    kwargs = build_embed_option_kwargs(
+        "https://litellm.example.com/v1/embeddings",
+        "mistral/embed-small",
+        embed_model_provider_prefix="acme",
+    )
+
+    assert kwargs["model_name"] == "acme/mistral/embed-small"
+    assert kwargs["embed_model_name"] == "acme/mistral/embed-small"
+
+
+def test_build_embed_option_kwargs_prefix_supports_bare_model_name():
+    kwargs = build_embed_option_kwargs(
+        "https://litellm.example.com/v1/embeddings",
+        "nv-embedqa-e5-v5",
+        embed_model_provider_prefix="nvidia",
+    )
+
+    assert kwargs["model_name"] == "nvidia/nv-embedqa-e5-v5"
+    assert kwargs["embed_model_name"] == "nvidia/nv-embedqa-e5-v5"
+
+
+def test_build_embed_option_kwargs_prefix_is_remote_only():
+    kwargs = build_embed_option_kwargs(
+        None,
+        "nvidia/llama-nemotron-embed-vl-1b-v2",
+        embed_model_provider_prefix="nvidia",
+    )
+
+    assert kwargs["model_name"] == "nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert kwargs["embed_model_name"] == "nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert "embed_model_provider_prefix" not in kwargs
+
+
 # ===================================================================
 # embed_granularity
 # ===================================================================
diff --git a/nemo_retriever/tests/test_root_cli_workflow.py b/nemo_retriever/tests/test_root_cli_workflow.py
index f0decca4aa..659904565b 100644
--- a/nemo_retriever/tests/test_root_cli_workflow.py
+++ b/nemo_retriever/tests/test_root_cli_workflow.py
@@ -434,6 +434,8 @@ def fake_create_ingestor(**_kwargs: Any) -> Any:
             "http://embed:8000/v1/embeddings",
             "--embed-model-name",
             "nvidia/llama-nemotron-embed-1b-v2",
+            "--embed-model-provider-prefix",
+            "nvidia",
         ],
     )
 
@@ -450,8 +452,8 @@ def fake_create_ingestor(**_kwargs: Any) -> Any:
     assert isinstance(embed_params, EmbedParams)
     assert embed_params.embed_invoke_url == "http://embed:8000/v1/embeddings"
     assert embed_params.embedding_endpoint == "http://embed:8000/v1/embeddings"
-    assert embed_params.model_name == "nvidia/llama-nemotron-embed-1b-v2"
-    assert embed_params.embed_model_name == "nvidia/llama-nemotron-embed-1b-v2"
+    assert embed_params.model_name == "nvidia/nvidia/llama-nemotron-embed-1b-v2"
+    assert embed_params.embed_model_name == "nvidia/nvidia/llama-nemotron-embed-1b-v2"
 
 
 def test_root_ingest_passes_migrated_extraction_and_embedding_flags(monkeypatch, tmp_path) -> None:
diff --git a/nemo_retriever/tests/test_root_query_cli.py b/nemo_retriever/tests/test_root_query_cli.py
index 28bfeec9bf..f66a86f6aa 100644
--- a/nemo_retriever/tests/test_root_query_cli.py
+++ b/nemo_retriever/tests/test_root_query_cli.py
@@ -136,6 +136,8 @@ def query(self, query: str, **_kwargs: Any) -> list[dict[str, Any]]:
             "http://embed:8000/v1/embeddings",
             "--embed-model-name",
             "nvidia/llama-nemotron-embed-1b-v2",
+            "--embed-model-provider-prefix",
+            "nvidia",
         ],
     )
 
@@ -148,8 +150,8 @@ def query(self, query: str, **_kwargs: Any) -> list[dict[str, Any]]:
             "embed_kwargs": {
                 "embed_invoke_url": "http://embed:8000/v1/embeddings",
                 "embedding_endpoint": "http://embed:8000/v1/embeddings",
-                "model_name": "nvidia/llama-nemotron-embed-1b-v2",
-                "embed_model_name": "nvidia/llama-nemotron-embed-1b-v2",
+                "model_name": "nvidia/nvidia/llama-nemotron-embed-1b-v2",
+                "embed_model_name": "nvidia/nvidia/llama-nemotron-embed-1b-v2",
             },
         }
     ]
diff --git a/nemo_retriever/tests/test_service_local_models.py b/nemo_retriever/tests/test_service_local_models.py
index e7abd6e3ba..fae41ed351 100644
--- a/nemo_retriever/tests/test_service_local_models.py
+++ b/nemo_retriever/tests/test_service_local_models.py
@@ -27,6 +27,7 @@ def test_build_embed_params_from_nim_config() -> None:
     nim = NimEndpointsConfig(
         embed_invoke_url="http://embed-nim/v1/embeddings",
         embed_model_name="nvidia/llama-nemotron-embed-vl-1b-v2",
+        embed_model_provider_prefix="nvidia",
         api_key="k",
     )
     ep = build_embed_params(nim, LocalModelsConfig(enabled=True))
@@ -34,6 +35,7 @@ def test_build_embed_params_from_nim_config() -> None:
     assert ep.embed_invoke_url == "http://embed-nim/v1/embeddings"
     assert ep.model_name == "nvidia/llama-nemotron-embed-vl-1b-v2"
     assert ep.embed_model_name == "nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert ep.embed_model_provider_prefix == "nvidia"
     assert ep.api_key == "k"
 
 
diff --git a/nemo_retriever/tests/test_service_vectordb_app.py b/nemo_retriever/tests/test_service_vectordb_app.py
index 1ea0a12d55..04b0e54c92 100644
--- a/nemo_retriever/tests/test_service_vectordb_app.py
+++ b/nemo_retriever/tests/test_service_vectordb_app.py
@@ -10,6 +10,7 @@
 
 from nemo_retriever.service.vectordb_app import (
     VectorDBState,
+    _embed_queries_remote,
     _tensor_to_embedding_rows,
     create_vectordb_app,
 )
@@ -84,6 +85,31 @@ def test_vector_db_state_local_embed_queries() -> None:
     mock_embedder.embed_queries.assert_called_once_with(["hello"])
 
 
+def test_remote_embed_queries_delegates_model_prefix(monkeypatch) -> None:
+    calls = {}
+
+    def fake_infer_microservice(data, **kwargs):
+        calls["data"] = data
+        calls.update(kwargs)
+        return [[0.1, 0.2]]
+
+    monkeypatch.setattr("nemo_retriever.models.nim.util.infer_microservice", fake_infer_microservice)
+
+    vectors = _embed_queries_remote(
+        ["hello"],
+        embed_model="nvidia/llama-nemotron-embed-vl-1b-v2",
+        embed_endpoint="https://litellm.example.com/v1/embeddings",
+        embed_api_key="k",
+        embed_model_provider_prefix="nvidia",
+    )
+
+    assert vectors == [[0.1, 0.2]]
+    assert calls["data"] == ["hello"]
+    assert calls["model_name"] == "nvidia/llama-nemotron-embed-vl-1b-v2"
+    assert calls["model_provider_prefix"] == "nvidia"
+    assert calls["embedding_endpoint"] == "https://litellm.example.com/v1/embeddings"
+
+
 _CANNED_HITS = [
     {
         "text": "Revenue grew 12% year over year.",