[riku_iki]

> That's not the same as starting with open model weights and training on those - until recently (gpt-oss) OpenAI didn't release their model weights.

there was obviously llama.