From 7fec1b58ea27d0cb93d49f21f4ed61a1da5b5eda Mon Sep 17 00:00:00 2001 From: Sipan Petrosyan Date: Wed, 29 Apr 2026 08:23:45 +0100 Subject: [PATCH] fix: clamp max_output_tokens to OpenAI min of 16 in anthropic-to-responses adapter --- .../responses_adapters/transformation.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py index 2badc2a3276..72a718c6300 100644 --- a/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py +++ b/litellm/llms/anthropic/experimental_pass_through/responses_adapters/transformation.py @@ -333,8 +333,10 @@ class LiteLLMAnthropicToResponsesAPIAdapter: responses_kwargs["instructions"] = "\n".join(filter(None, text_parts)) # max_tokens -> max_output_tokens + # OpenAI Responses API requires max_output_tokens >= 16 max_tokens = anthropic_request.get("max_tokens") if max_tokens: + max_tokens = max(max_tokens, 16) responses_kwargs["max_output_tokens"] = max_tokens # temperature / top_p passed through