|
|
This post was updated on .
ChatGPT is NOT AN AI MODEL!! IT IS AN INTERFACE FOR INTERACTING WITH OPENAI'S INSTANCE OF THE GPT MODEL.
I use all the models—GPT, Claude, Grok, Gemini, Sonar (via Perplexity, although Sonar is technically more of a RAG index than an agential model), Kimi K2, etc. From my experience, Claude generates the best textual content, Grok is the most tolerant of political incorrectness, and GPT has the most accurate voice transcription.
Kimi K2 is interesting and unique, although it is extremely woke. Since it is a Chinese model, I am thoroughly disappointed by its wokeness. MiniMax and Moonshot AI (Kimi's developer) appears to have trained heavily on Western RLHF frameworks and datasets, possibly to compete in Western markets. This is really fucking annoying. Nonetheless, Kimi K2 is particularly useful for deep research.
|