None of these tests involve actual intelligent discussion. As a user of Anthropic's Claude, I was extremely disappointed with DeepSeek. It had to be reminded about things which were said earlier in the chat, and even then it seemed to be faking the contextual awareness.