2 ms·
Makes sense as claude instant is likely better than 3.5
by whimsicalism 3y ago
Makes sense as claude instant is likely better than 3.5
- phillipcarter 3y agoI dunno about that. GPT 3.5 is extremely good. I would wager that most apps that use RAG to pass context in and get JSON (or some other thing) out that you can pass to some other part of your product don't need GPT 4 or anything else equally as powerful.
- whimsicalism 3y ago> GPT 3.5 is extremely good Maybe I just use GPT4 too much, but I disagree and most benchmarks show Clause being neck-and-neck with 3.5, especially the lmsys benchmarks which I think are the highest quality. [0] MMLU is basically broken (although even that puts Claude higher). [0]: https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboar...