Whenever a domestic large model releases something, they always get Arena to run a benchmark score, and then it’s the classic “massively beating Claude and GPT.” Is it always like this? Either it’s a frontend benchmark, or some specialized task benchmark—anyway, every time they post it, it’s like they’re trying to make a big news story.

At first, I even went out of my way to try it—just to test it. Later, I couldn’t be bothered to try anymore. It’s like how domestic companies used to benchmark themselves to death and “crush” Apple—what’s the point, really……

Arena is basically just “collect money and do the job.” Just treat it as entertainment. If you don’t use the ready-made Fable Sol or ignore SOTA, and instead go use other models, that’s just wasting time.