Add DeepSeek thinking/tool-call guidance with reasoning_content

[Harrrryz]

Title

Add DeepSeek thinking/tool-call guidance with reasoning_content

Relevant issues

N/A (docs-only guidance for DeepSeek thinking mode)

Pre-Submission checklist

• I have Added testing in the tests/litellm/ (https://github.com/BerriAI/litellm/tree/main/tests/litellm) directory, Adding at least 1 test is a hard requirement - see details (https://docs.litellm.ai/docs/extras/contributing_code)
• I have added a screenshot of my new test passing locally
• My PR passes all unit tests on make test-unit (https://docs.litellm.ai/docs/extras/contributing_code)
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

📖 Documentation

Changes

• Update DeepSeek provider docs with a “Tool calling with thinking / reasoning_content” section explaining that DeepSeek requires echoing reasoning_content on assistant tool-call messages when thinking is enabled, linking to the official guide.
• Add a full Python example showing how to enable thinking, replay reasoning_content, send tool outputs, and optionally strip reasoning content before a new user turn.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
litellm	Ready	Preview	Comment	Dec 8, 2025 3:13am

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[stonyme]

thank you ,We need it.

[krrishdholakia]

why is this in extra body? can't we support this through the openai-compatible 'reasoning_effort' parameter or the anthropic thinking parameter? https://docs.litellm.ai/docs/reasoning_content#pass-thinking-to-anthropic-models

[krrishdholakia]

cc: @Chesars

[Chesars]

Yes, extra_body works as a pass-through for provider-specific params, but we should support this param natively in a open-ai/anthropic format. PR #17712 adds it

[Chesars]

Suggested change

	extra_body={"thinking": {"type": "enabled"}},
	thinking={"type": "enabled"}

[Chesars]

Suggested change

	extra_body={"thinking": {"type": "enabled"}},
	thinking={"type": "enabled"}