Is it just me or Kimi K3.1 Pro don’t look much better than DeepSeek 4.1 Flash?
Personally I am quite fond of the Kimi models, especially them taking a gamble pushing the 2.8b parameter models. But that’s just it, you’d figure that the results would be representative of its size. However, the recent release of DeepSeek 1017 has put things in perspective for me.