diff --git a/litellm/llms/anthropic/chat/handler.py b/litellm/llms/anthropic/chat/handler.py
index ef0f45d8f8b..822f859b9e2 100644
--- a/litellm/llms/anthropic/chat/handler.py
+++ b/litellm/llms/anthropic/chat/handler.py
@@ -845,7 +845,10 @@ class ModelResponseIterator:
                 # Track current content block type for filtering deltas
                 self.current_content_block_type = content_block_start["content_block"]["type"]
                 if content_block_start["content_block"]["type"] == "text":
-                    text = content_block_start["content_block"]["text"]
+                    # Anthropic always sends "text": "" here, but some compatible gateways omit
+                    # the key. Absence on a text block start has the same meaning as the empty
+                    # string, so read it leniently rather than dropping the stream.
+                    text = content_block_start["content_block"].get("text", "")
                 elif (
                     content_block_start["content_block"]["type"] == "tool_use"
                     or content_block_start["content_block"]["type"] == "server_tool_use"
diff --git a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py
index 1e0d2e55373..fa5f3edd23a 100644
--- a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py
+++ b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py
@@ -533,6 +533,42 @@ def test_regular_tool_finish_reason():
     assert model_response.choices[0].finish_reason == "tool_calls"
 
 
+def test_content_block_start_without_text_key_does_not_kill_the_stream():
+    """Some Anthropic-compatible gateways omit "text" on a text block start.
+
+    api.anthropic.com always sends "text": "", so absence carries the same meaning.
+    https://github.com/BerriAI/litellm/issues/40689
+    """
+    chunks = [
+        {
+            "type": "message_start",
+            "message": {
+                "id": "msg_123",
+                "type": "message",
+                "role": "assistant",
+                "content": [],
+                "usage": {"input_tokens": 10, "output_tokens": 1},
+            },
+        },
+        # no "text" key, unlike the official API
+        {"type": "content_block_start", "index": 0, "content_block": {"type": "text"}},
+        {
+            "type": "content_block_delta",
+            "index": 0,
+            "delta": {"type": "text_delta", "text": "hi"},
+        },
+        {"type": "content_block_stop", "index": 0},
+    ]
+
+    iterator = ModelResponseIterator(None, sync_stream=True)
+    parsed = [iterator.chunk_parser(chunk) for chunk in chunks]
+
+    start = parsed[1]
+    assert start.choices[0].delta.content == ""
+    delta = parsed[2]
+    assert delta.choices[0].delta.content == "hi"
+
+
 def test_text_only_streaming_has_index_zero():
     """Test that text-only streaming responses have choice index=0"""
     chunks = [