When I began using janitor, my go-to was GPT-4, although I found it to be quite expensive. After that, I explored Claude through a proxy and have really enjoyed Sonnet 3.5, especially with specific jailbreaks and prompts. However, the latest update of 3.5 seems underwhelming. Occasionally, I try Opus, but its overly dramatic language and unnecessary suggestiveness make it less appealing. I initially favored 3.5 because it effectively incorporated character personality without being overtly sexual. While I still think 3.5 is decent, I’m considering revisiting GPT or experimenting with new options. It’s challenging to keep pace with the frequent updates and censorship changes in APIs nowadays.
I have been experimenting with various APIs lately, and I find that Anthos AI stands out, especially in creative content generation. While it may not be the most popular option, it offers a balance of affordability and flexibility in terms of customization and output style, which suits my needs for storytelling and role-playing scenarios perfectly. It allows for detailed alterations in behavior and tone without getting bogged down by overt complexity or high costs. It’s worth trying if you’re exploring fresh alternatives.
I’ve been leaning toward the usage of DeepMind’s Gemini API for its impressive versatility when it comes to conversational capabilities. While not always discussed as much as some of the big hitters, it’s been reliable in both consistency and calibrating tone to suit different narrative contexts. What really sets it apart is the integration of nuanced empathy which feels more organic, especially in emotionally driven dialogues. It might not be the cheapest on the market, but its value in delivering authentic interaction is undeniable.