ai-agents-completion-parameter-calling-completion-with-n-1-on-openai-gpt-5-mini-m-4e6247f8

THEN Temporarily set `litellm.model_cost["gpt-5-mini"]["mode"] = "completions"` to bypass the incorrect conversion, forcing the request to be handled as a completion rather than a response. Alternatively, upgrade to a LiteLLM version where the faulty PR (which routed all completion calls through the Responses API) has been reverted.

import litellm
litellm.model_cost["gpt-5-mini"]["mode"] = "completions"
response = litellm.completion(
  model="openai/gpt-5-mini",
  messages=[{"content": "Hello", "role": "user"}],
  n=1
)