Is reasoning content revealed in prompt playground?

[qlibp]

I don't see any reasoning content in trace. I'm currently runing deepseek-reasoner model.

[dosubot[bot]]

Reasoning content in the prompt playground is generally shown as part of the assistant's response text, especially when using models like deepseek-reasoner that follow the OpenAI-compatible API. The playground displays whatever the model returns—so if the model's output includes step-by-step reasoning, it should appear in the chat output or trace. There isn't a special UI for "reasoning" with deepseek-reasoner; it's just part of the assistant's message content, unlike Anthropic models which have a dedicated reasoning config and UI.

If you don't see any reasoning in the trace, it's likely because the model isn't returning explicit reasoning steps in its output, or the prompt isn't structured to elicit that kind of response. You might want to try using ReAct-style prompts or explicitly instructing the model to "think step-by-step" in your prompt to encourage more reasoning in the output. For more on how reasoning is surfaced with ReAct prompting, see the docs.

If you expect reasoning and still don't see it, double-check that the model is returning the expected content by inspecting the raw output in the playground. If the output is present but not visible, there could be a parsing or display issue, but the playground is designed to show all assistant responses for deepseek-reasoner as plain text in the chat output.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[axiomofjoy]

Thanks @qlibp. Is this trace you're showing coming from a run of prompt playground or via an instrumented client application?

[qlibp]

from prompt playground

[axiomofjoy]

Thanks @qlibp, I've filed an issue here.