This blog post announces the release of llm 0.32a2, highlighting a significant change in OpenAI's reasoning-capable models that now use the /v1/responses endpoint, allowing for improved reasoning display in outputs. It includes functionality for hiding summarized reasoning tokens.