How to set reasoning effort via system message for GPT-OSS (20B) in llama.cpp?

[antonkratz]

I am running OpenAI's gpt-oss:20B with llama.cpp. I want to supply a system message to adjust the reasoning effort. How can I do that?

From https://openai.com/index/introducing-gpt-oss/: "Similar to the OpenAI o-series reasoning models in the API, the two open-weight models support three reasoning efforts—low, medium, and high—which trade off latency vs. performance. Developers can easily set the reasoning effort with one sentence in the system message."