[levihanlen]

Currently the model lineup includes:

- Gemini 2.0 Flash. This is what I use most often, as the quality/price ratio is amazing. (Looking into 2.5 flash preview)

- Deepseek v3 03.24

- GPT 4.1 (The most expensive model by far. The rest are very cheap. About 20x more expensive that 2.0 flash for my use cases)

- GPT 4.1 mini

- Qwen 3 235B and Qwen 3 30B MoE

- Grok 3 mini

- Gemma 3 27B

[kadushka]

I haven’t tried any of these, recently I’ve been switching between Claude 3.7, Gemini 2.5 pro, and GPT-4.5, with each one producing something interesting once in a while. I usually sketch a plot (sci-fi short story: 10-20 pages), mention a few details of the environment, describe the main character (often based on myself), and mention relevant writing styles. Sometimes I describe how I want it to write the story, sometimes I don’t. I rarely interact with the story beyond the initial prompt, it’s easier to just resubmit if I don’t like something.

[levihanlen]

Those are some great models. The reason I don't use them is simply the cost (especially 4.5).

And that's a pretty cool process you use!