Accessing reasoning tokens of another llm model in agents sdk

[atahanozdemirberkeley]

I'm using the agents SDK with a non-OpenAI model that supports reasoning (using a ":thinking" suffix in the model name).

When using these models, I can see that the Generation output includes a count of "reasoning_tokens" in the usage stats:

"usage": {
  "input_tokens": 13545,
  "input_tokens_details": { "cached_tokens": 0 },
  "output_tokens_details": { "reasoning_tokens": 114 },
  "output_tokens": 270,
  "total_tokens": 13815
}

However, I can't find a way to access the actual reasoning content. For OpenAI models, I understand the SDK supports reasoning configuration in ModelSettings to expose reasoning content, but that is only supported for o-series models it seems like.

Is there any way to access the reasoning content from non-OpenAI models through the Agents SDK, or is this currently only supported for OpenAI models?

[paulsengh]

Hey @rm-openai any thoughts here? Thank you!

[github-actions]

This issue is stale because it has been open for 7 days with no activity.

[rm-openai]

Sorry missed this. Can you provide some sample code with the different model?

[github-actions]

This issue is stale because it has been open for 7 days with no activity.

[github-actions]

This issue was closed because it has been inactive for 3 days since being marked as stale.