[Bedrock] service_tier rejected for openai.gpt-oss-120b-1:0 unless allowed_openai_params is manually set

[kuldeepjain]

What happened? (Also tell us what you expected to happen)

When calling the Bedrock model openai.gpt-oss-120b-1:0 using the Chat Completions API, LiteLLM rejects the service_tier parameter with an UnsupportedParamsError.

Expected behavior:
LiteLLM should automatically forward service_tier to Bedrock for models that support AWS Service Tiers, without requiring the caller to explicitly whitelist it through allowed_openai_params.

AWS Bedrock does support service tiers for this model:

• https://docs.aws.amazon.com/bedrock/latest/userguide/service-tiers-inference.html
• https://aws.amazon.com/bedrock/service-tiers/

---

❌ Request Payload (fails)

{
  "messages": [
    { "role": "user", "content": "Hi" }
  ],
  "max_tokens": 100,
  "service_tier": "priority"
}

Response – 400 Bad Request

UnsupportedParamsError: bedrock does not support parameters: ['service_tier'], for model=openai.gpt-oss-120b-1:0.
To drop these, set drop_params=True or for proxy: litellm_settings: drop_params: true.
If you want to use these params dynamically send allowed_openai_params=['service_tier'] in your request.

---

✅ Request Payload (works only when manually whitelisting)

{
  "messages": [
    { "role": "user", "content": "Hi" }
  ],
  "max_tokens": 100,
  "service_tier": "priority",
  "allowed_openai_params": ["service_tier"]
}

Response Payload

{
  "id": "chatcmpl-b402c763-fda9-42bb-9894-84e5f2351662",
  "model": "openai.gpt-oss-120b-1:0",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "",
        "reasoning_content": "The user says \"Hi\". Lik",
        "thinking_blocks": [
          { "type": "thinking", "thinking": "The user says \"Hi\". Lik" }
        ]
      }
    }
  ]
}

---

Steps to Reproduce

1. Call Bedrock model openai.gpt-oss-120b-1:0 with a standard Chat Completions payload that includes "service_tier": "priority"
2. Observe UnsupportedParamsError
3. Add "allowed_openai_params": ["service_tier"]
4. Observe that the request now succeeds

---

Expected Behavior

• LiteLLM should automatically allow service_tier for supported Bedrock models
• The caller should not need to manually set allowed_openai_params
• The parameter should not be treated as unsupported when AWS Bedrock accepts it

---

Relevant Log Output

  File "/app/app/service/v3/services/chat_service.py", line 541, in _acompletion_with_retry
    return await acompletion(**params)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/utils.py", line 1637, in wrapper_async
    raise e
  File "/app/.venv/lib/python3.11/site-packages/litellm/utils.py", line 1483, in wrapper_async
    result = await original_function(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/main.py", line 614, in acompletion
    raise exception_type(
  File "/app/.venv/lib/python3.11/site-packages/litellm/main.py", line 587, in acompletion
    init_response = await loop.run_in_executor(None, func_with_context)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/utils.py", line 1098, in wrapper
    result = original_function(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/main.py", line 3767, in completion
    raise exception_type(
  File "/app/.venv/lib/python3.11/site-packages/litellm/main.py", line 1318, in completion
    optional_params = get_optional_params(
                      ^^^^^^^^^^^^^^^^^^^^
  File "/app/.venv/lib/python3.11/site-packages/litellm/utils.py", line 3469, in get_optional_params
    _check_valid_arg(
  File "/app/.venv/lib/python3.11/site-packages/litellm/utils.py", line 3452, in _check_valid_arg
    raise UnsupportedParamsError(
litellm.exceptions.UnsupportedParamsError: litellm.UnsupportedParamsError: bedrock does not support parameters: ['service_tier'], for model=openai.gpt-oss-120b-1:0. To drop these, set `litellm.drop_params=True` or for proxy:
`litellm_settings:
 drop_params: true`

Full logs included above.

---

LiteLLM version

1.79.1

---

Are you a ML Ops Team?

Yes
(We operate the ML Platform that exposes LLM models internally.)

[kuldeepjain]

Looks like this is fixed here: #17810
When is it getting released: https://pypi.org/project/litellm/#history ?