feat(sambanova): add MiniMax-M3, update model limits, support extra chat params

litellm/llms/sambanova/chat.py

@@ -11,8 +11,12 @@
     handle_messages_with_content_list_to_str_conversion,
 )
 from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
+from litellm.secret_managers.main import get_secret_str
 from litellm.types.llms.openai import AllMessageValues
 
+DEFAULT_INTEGRATION_SOURCE: Final = "litellm"
+INTEGRATION_SOURCE_HEADER: Final = "X-Integration-Source"
+
 
 class SambanovaConfig(OpenAIGPTConfig):
     """
@@ -59,9 +63,9 @@ def get_supported_openai_params(self, model: str) -> list:
         Get the supported OpenAI params for the given model
 
         """
-        from litellm.utils import supports_function_calling
+        from litellm.utils import supports_function_calling, supports_reasoning
 
-        params: Final = [
+        params: Final = (
             "max_completion_tokens",
             "max_tokens",
             "response_format",
@@ -71,14 +75,24 @@ def get_supported_openai_params(self, model: str) -> list:
             "temperature",
             "top_p",
             "top_k",
-        ]
-
-        if supports_function_calling(model, custom_llm_provider="sambanova"):
-            params.append("tools")
-            params.append("tool_choice")
-            params.append("parallel_tool_calls")
-
-        return params
+            "presence_penalty",
+            "frequency_penalty",
+            "logprobs",
+            "top_logprobs",
+            "n",
+            "logit_bias",
+            "seed",
+        )
+        tool_params: Final = (
+            ("tools", "tool_choice", "parallel_tool_calls")
+            if supports_function_calling(model, custom_llm_provider="sambanova")
+            else ()
+        )
+        reasoning_params: Final = (
+            ("reasoning_effort",) if supports_reasoning(model, custom_llm_provider="sambanova") else ()
+        )
+
+        return [*params, *tool_params, *reasoning_params]
 
     def map_openai_params(
         self,
@@ -98,6 +112,43 @@ def map_openai_params(
                 optional_params[param] = value
         return optional_params
 
+    def transform_request(
+        self,
+        model: str,
+        messages: list[AllMessageValues],
+        optional_params: dict,
+        litellm_params: dict,
+        headers: dict,
+    ) -> dict:
+        """Send X-Integration-Source: caller header, then extra_body, then env var, then default"""
+        extra_headers: Final[dict[str, str]] = dict(optional_params.get("extra_headers") or {})
+        extra_body: Final[dict[str, object]] = dict(optional_params.get("extra_body") or {})
+        body_source: Final = extra_body.pop("integration_source", None)
+        header_source: Final = next(
+            (value for key, value in extra_headers.items() if key.lower() == INTEGRATION_SOURCE_HEADER.lower()),
+            None,
+        )
+        integration_source: Final = (
+            body_source or get_secret_str("SAMBANOVA_INTEGRATION_SOURCE") or DEFAULT_INTEGRATION_SOURCE
+        )
+        headers_with_source: Final = (
+            extra_headers if header_source else {**extra_headers, INTEGRATION_SOURCE_HEADER: integration_source}
+        )
+        other_params: Final[dict[str, object]] = {k: v for k, v in optional_params.items() if k != "extra_body"}
+        updated_params: Final[dict[str, object]] = {
+            **other_params,
+            "extra_headers": headers_with_source,
+            **({"extra_body": extra_body} if extra_body else {}),
+        }
+
+        return super().transform_request(
+            model=model,
+            messages=messages,
+            optional_params=updated_params,
+            litellm_params=litellm_params,
+            headers=headers,
+        )
+
     @overload
     def _transform_messages(
         self, messages: list[AllMessageValues], model: str, is_async: Literal[True]

litellm/model_prices_and_context_window_backup.json

@@ -45768,39 +45768,26 @@
         "supports_reasoning": true,
         "supports_tool_choice": true
     },
-    "sambanova/DeepSeek-R1": {
-        "input_cost_per_token": 5e-06,
-        "litellm_provider": "sambanova",
-        "max_input_tokens": 32768,
-        "max_output_tokens": 32768,
-        "max_tokens": 32768,
-        "mode": "chat",
-        "output_cost_per_token": 7e-06,
-        "source": "https://cloud.sambanova.ai/plans/pricing"
-    },
-    "sambanova/Llama-4-Maverick-17B-128E-Instruct": {
-        "input_cost_per_token": 6.3e-07,
+    "sambanova/MiniMax-M3": {
+        "input_cost_per_token": 6e-07,
         "litellm_provider": "sambanova",
-        "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
-        "max_tokens": 131072,
-        "metadata": {
-            "notes": "For vision models, images are converted to 6432 input tokens and are billed at that amount"
-        },
+        "max_input_tokens": 1048576,
+        "max_output_tokens": 1048576,
+        "max_tokens": 1048576,
         "mode": "chat",
-        "output_cost_per_token": 1.8e-06,
+        "output_cost_per_token": 2.4e-06,
         "source": "https://cloud.sambanova.ai/plans/pricing",
         "supports_function_calling": true,
+        "supports_reasoning": true,
         "supports_response_schema": true,
-        "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_tool_choice": true
     },
     "sambanova/Meta-Llama-3.3-70B-Instruct": {
         "input_cost_per_token": 6e-07,
         "litellm_provider": "sambanova",
         "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
-        "max_tokens": 131072,
+        "max_output_tokens": 3072,
+        "max_tokens": 3072,
         "mode": "chat",
         "output_cost_per_token": 1.2e-06,
         "source": "https://cloud.sambanova.ai/plans/pricing",
@@ -45809,9 +45796,9 @@
         "supports_tool_choice": true
     },
     "sambanova/DeepSeek-V3.1": {
-        "max_tokens": 131072,
+        "max_tokens": 7168,
         "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
+        "max_output_tokens": 7168,
         "input_cost_per_token": 3e-06,
         "output_cost_per_token": 4.5e-06,
         "litellm_provider": "sambanova",
@@ -45835,9 +45822,9 @@
         "source": "https://cloud.sambanova.ai/plans/pricing"
     },
     "sambanova/DeepSeek-V3.2": {
-        "max_tokens": 32768,
+        "max_tokens": 7168,
         "max_input_tokens": 32768,
-        "max_output_tokens": 32768,
+        "max_output_tokens": 7168,
         "input_cost_per_token": 3e-06,
         "output_cost_per_token": 4.5e-06,
         "litellm_provider": "sambanova",

model_prices_and_context_window.json

@@ -45768,39 +45768,26 @@
         "supports_reasoning": true,
         "supports_tool_choice": true
     },
-    "sambanova/DeepSeek-R1": {
-        "input_cost_per_token": 5e-06,
-        "litellm_provider": "sambanova",
-        "max_input_tokens": 32768,
-        "max_output_tokens": 32768,
-        "max_tokens": 32768,
-        "mode": "chat",
-        "output_cost_per_token": 7e-06,
-        "source": "https://cloud.sambanova.ai/plans/pricing"
-    },
-    "sambanova/Llama-4-Maverick-17B-128E-Instruct": {
-        "input_cost_per_token": 6.3e-07,
+    "sambanova/MiniMax-M3": {
+        "input_cost_per_token": 6e-07,
         "litellm_provider": "sambanova",
-        "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
-        "max_tokens": 131072,
-        "metadata": {
-            "notes": "For vision models, images are converted to 6432 input tokens and are billed at that amount"
-        },
+        "max_input_tokens": 1048576,
+        "max_output_tokens": 1048576,
+        "max_tokens": 1048576,
         "mode": "chat",
-        "output_cost_per_token": 1.8e-06,
+        "output_cost_per_token": 2.4e-06,
         "source": "https://cloud.sambanova.ai/plans/pricing",
         "supports_function_calling": true,
+        "supports_reasoning": true,
         "supports_response_schema": true,
-        "supports_tool_choice": true,
-        "supports_vision": true
+        "supports_tool_choice": true
     },
     "sambanova/Meta-Llama-3.3-70B-Instruct": {
         "input_cost_per_token": 6e-07,
         "litellm_provider": "sambanova",
         "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
-        "max_tokens": 131072,
+        "max_output_tokens": 3072,
+        "max_tokens": 3072,
         "mode": "chat",
         "output_cost_per_token": 1.2e-06,
         "source": "https://cloud.sambanova.ai/plans/pricing",
@@ -45809,9 +45796,9 @@
         "supports_tool_choice": true
     },
     "sambanova/DeepSeek-V3.1": {
-        "max_tokens": 131072,
+        "max_tokens": 7168,
         "max_input_tokens": 131072,
-        "max_output_tokens": 131072,
+        "max_output_tokens": 7168,
         "input_cost_per_token": 3e-06,
         "output_cost_per_token": 4.5e-06,
         "litellm_provider": "sambanova",
@@ -45835,9 +45822,9 @@
         "source": "https://cloud.sambanova.ai/plans/pricing"
     },
     "sambanova/DeepSeek-V3.2": {
-        "max_tokens": 32768,
+        "max_tokens": 7168,
         "max_input_tokens": 32768,
-        "max_output_tokens": 32768,
+        "max_output_tokens": 7168,
         "input_cost_per_token": 3e-06,
         "output_cost_per_token": 4.5e-06,
         "litellm_provider": "sambanova",

tests/unit/llms/sambanova/test_sambanova_chat_transformation.py

@@ -0,0 +1,127 @@
+import pytest
+
+from litellm.llms.sambanova.chat import SambanovaConfig
+from litellm.utils import get_optional_params
+
+_ALWAYS_SUPPORTED = (
+    "max_tokens",
+    "max_completion_tokens",
+    "temperature",
+    "top_p",
+    "top_k",
+    "stop",
+    "stream",
+    "stream_options",
+    "response_format",
+    "presence_penalty",
+    "frequency_penalty",
+    "logprobs",
+    "top_logprobs",
+    "n",
+    "logit_bias",
+    "seed",
+)
+
+
+@pytest.mark.parametrize("param", _ALWAYS_SUPPORTED)
+def test_supported_params_include_spec_params(param):
+    assert param in SambanovaConfig().get_supported_openai_params("MiniMax-M3")
+
+
+def test_reasoning_effort_only_for_reasoning_models():
+    config = SambanovaConfig()
+    assert "reasoning_effort" in config.get_supported_openai_params("gpt-oss-120b")
+    assert "reasoning_effort" not in config.get_supported_openai_params("Meta-Llama-3.3-70B-Instruct")
+
+
+def test_tools_only_for_function_calling_models():
+    config = SambanovaConfig()
+    for param in ("tools", "tool_choice", "parallel_tool_calls"):
+        assert param in config.get_supported_openai_params("MiniMax-M3")
+        assert param not in config.get_supported_openai_params("gemma-4-31B-it")
+
+
+def test_max_completion_tokens_mapped_to_max_tokens():
+    optional_params = SambanovaConfig().map_openai_params(
+        non_default_params={"max_completion_tokens": 128},
+        optional_params={},
+        model="MiniMax-M3",
+        drop_params=False,
+    )
+    assert optional_params == {"max_tokens": 128}
+
+
+def test_new_params_forwarded_via_get_optional_params():
+    optional_params = get_optional_params(
+        model="MiniMax-M3",
+        custom_llm_provider="sambanova",
+        seed=7,
+        presence_penalty=0.5,
+        frequency_penalty=0.25,
+        reasoning_effort="low",
+    )
+    assert optional_params["seed"] == 7
+    assert optional_params["presence_penalty"] == 0.5
+    assert optional_params["frequency_penalty"] == 0.25
+    assert optional_params["reasoning_effort"] == "low"
+
+
+def _transform(optional_params):
+    return SambanovaConfig().transform_request(
+        model="MiniMax-M3",
+        messages=[{"role": "user", "content": "hi"}],
+        optional_params=optional_params,
+        litellm_params={},
+        headers={},
+    )
+
+
+def test_integration_source_defaults_to_litellm(monkeypatch):
+    monkeypatch.delenv("SAMBANOVA_INTEGRATION_SOURCE", raising=False)
+    request = _transform({})
+    assert request["extra_headers"]["X-Integration-Source"] == "litellm"
+
+
+def test_integration_source_env_overrides_default(monkeypatch):
+    monkeypatch.setenv("SAMBANOVA_INTEGRATION_SOURCE", "langflow")
+    request = _transform({})
+    assert request["extra_headers"]["X-Integration-Source"] == "langflow"
+
+
+def test_integration_source_param_overrides_env(monkeypatch):
+    monkeypatch.setenv("SAMBANOVA_INTEGRATION_SOURCE", "langflow")
+    optional_params = {"extra_body": {"integration_source": "my-framework"}}
+    request = _transform(optional_params)
+    assert request["extra_headers"]["X-Integration-Source"] == "my-framework"
+    assert "integration_source" not in request.get("extra_body", {})
+    assert "integration_source" in optional_params["extra_body"], "caller's dict must not be mutated"
+
+
+def test_integration_source_explicit_header_wins(monkeypatch):
+    monkeypatch.setenv("SAMBANOVA_INTEGRATION_SOURCE", "langflow")
+    request = _transform({"extra_headers": {"X-Integration-Source": "custom", "X-Other": "1"}})
+    assert request["extra_headers"] == {"X-Integration-Source": "custom", "X-Other": "1"}
+
+
+def test_integration_source_removed_from_body_when_header_also_set():
+    request = _transform(
+        {
+            "extra_headers": {"X-Integration-Source": "custom"},
+            "extra_body": {"integration_source": "ignored", "other": 1},
+        }
+    )
+    assert request["extra_headers"] == {"X-Integration-Source": "custom"}
+    assert request["extra_body"] == {"other": 1}
+
+
+def test_integration_source_header_matched_case_insensitively(monkeypatch):
+    monkeypatch.setenv("SAMBANOVA_INTEGRATION_SOURCE", "langflow")
+    request = _transform({"extra_headers": {"x-integration-source": "custom"}})
+    assert request["extra_headers"] == {"x-integration-source": "custom"}
+
+
+def test_content_list_converted_to_string():
+    messages = [{"role": "user", "content": [{"type": "text", "text": "a"}, {"type": "text", "text": "b"}]}]
+    out = SambanovaConfig()._transform_messages(messages, model="MiniMax-M3")
+    assert isinstance(out[0]["content"], str)
+    assert "a" in out[0]["content"] and "b" in out[0]["content"]