diff --git a/litellm/llms/oci/chat/cohere.py b/litellm/llms/oci/chat/cohere.py
index 62cb71f2a52..f13a371c85c 100644
--- a/litellm/llms/oci/chat/cohere.py
+++ b/litellm/llms/oci/chat/cohere.py
@@ -254,11 +254,14 @@ def handle_cohere_stream_chunk(dict_chunk: dict) -> ModelResponseStream:
         typed_chunk.index = 0
 
     # OCI Cohere's terminal SSE event re-sends the full assembled response in
-    # `text` alongside a populated `chatHistory`. Emitting that text would
-    # concatenate the whole response onto the already-streamed deltas.
-    # `chatHistory` is the correct discriminator: `finishReason` is a weaker
-    # signal that could in principle appear on a non-consolidated chunk.
-    is_terminal_consolidation = typed_chunk.chatHistory is not None
+    # `text` alongside a populated `chatHistory` and a non-null `finishReason`.
+    # Emitting that text would concatenate the whole response onto the
+    # already-streamed deltas. We require both signals to be present so that a
+    # future API change which adds `chatHistory` to intermediate chunks (or a
+    # rare early-populated case) doesn't silently drop legitimate token deltas.
+    is_terminal_consolidation = (
+        typed_chunk.chatHistory is not None and typed_chunk.finishReason is not None
+    )
     text = "" if is_terminal_consolidation else (typed_chunk.text or "")
 
     finish_reason = typed_chunk.finishReason
diff --git a/litellm/llms/oci/chat/generic.py b/litellm/llms/oci/chat/generic.py
index 7938e8f3254..bf44ee9b82f 100644
--- a/litellm/llms/oci/chat/generic.py
+++ b/litellm/llms/oci/chat/generic.py
@@ -327,7 +327,7 @@ def handle_generic_response(
     reasoning_tokens: Optional[int] = None
     if (
         oci_usage.completionTokensDetails
-        and oci_usage.completionTokensDetails.reasoningTokens
+        and oci_usage.completionTokensDetails.reasoningTokens is not None
     ):
         reasoning_tokens = oci_usage.completionTokensDetails.reasoningTokens
     model_response.usage = Usage(  # type: ignore[attr-defined]
diff --git a/litellm/llms/oci/chat/transformation.py b/litellm/llms/oci/chat/transformation.py
index e03773397ff..d34a34855d1 100644
--- a/litellm/llms/oci/chat/transformation.py
+++ b/litellm/llms/oci/chat/transformation.py
@@ -195,11 +195,15 @@ class OCIChatConfig(BaseConfig):
         # - tool_choice is unsupported
         # - stop sequences key is "stopSequences" not "stop"
         # - n (numGenerations) is GENERIC-only
+        # The unsupported keys are kept in the map with value ``False`` so
+        # ``map_openai_params`` either drops them (under drop_params) or raises
+        # a clear error, rather than silently passing them through.
         self.openai_to_oci_cohere_param_map = {
             k: ("stopSequences" if k == "stop" else v)
             for k, v in self.openai_to_oci_generic_param_map.items()
-            if k not in ("tool_choice", "max_retries", "n")
         }
+        self.openai_to_oci_cohere_param_map["tool_choice"] = False
+        self.openai_to_oci_cohere_param_map["n"] = False
         # OCI Cohere models are not reasoning models; mark reasoning_effort
         # explicitly unsupported so callers either get a clear error or have
         # the param dropped under drop_params, rather than silently passing
@@ -316,7 +320,12 @@ class OCIChatConfig(BaseConfig):
         litellm_params: dict,
         stream: Optional[bool] = None,
     ) -> str:
-        base = get_oci_base_url(optional_params, api_base or litellm.api_base)
+        # If the caller provides a full endpoint URL, use it as-is.
+        # Otherwise construct the standard OCI GenAI chat endpoint from the region.
+        resolved_base = api_base or litellm.api_base
+        if resolved_base:
+            return resolved_base.rstrip("/")
+        base = get_oci_base_url(optional_params, None)
         return f"{base}/{OCI_API_VERSION}/actions/chat"
 
     def _get_optional_params(