nemesis on Nostr: for chatgpt I use gpt-4o as my general baseline as it's very fast and pretty good if ...

for chatgpt I use gpt-4o as my general baseline as it's very fast and pretty good if you give it a reasonable prompt. deep research is extremely good if you want a literature review of something (though btw gpt-4o can do web searches too). o3 is very good (by LLM standards) at reasoning-heavy tasks in my experience but it's overkill as a goto option. gpt-4.5 is afaik a bit more hallucination-resistant and feels less artificial on a vibe level but I personally havent noticed it being actually much better than gpt-4o in practice