ai-agents-llm-configuration-when-using-openai-gpt-5-chat-model-gpt-5-chat-late-6113528d

THEN Set litellm.drop_params=True to drop the unsupported temperature parameter, or use temperature=1. The allowed_openai_params workaround may not override this restriction; dropping unsupported params is the reliable fix.

import litellm
litellm.drop_params = True
# Now client.chat.completions.create(model='openai/gpt-5-chat-latest', temperature=0.3, ...) will drop temperature rather than raise an error.