Original caption
Replying to @Apsino I saw this on X and this response from OX Alpha is genuinely interesting. People are already calling the model trash or assuming they know exactly who made it, but we still don’t know who the provider is. The fact that it answered a politically sensitive question about Xi Jinping this directly also doesn’t fit what a lot of people expect from more heavily moderated Chinese-hosted models. That does not prove OX Alpha isn’t Chinese — different system prompts, hosting, and post-training can completely change how a model responds. But that’s exactly my point: imagine we find out Google, OpenAI, or some completely unexpected lab built this thing. People would immediately look at the same outputs differently. Judge the model by what it actually does. Test it. Find where it wins, find where it sucks, and stop pretending we know the answer before the provider is even revealed. [OX Alpha, OX Alpha AI, stealth AI model, Chinese AI models, GLM, Qwen, OpenRouter, AI censorship, AI reasoning model] #oxalpha #greenscreen