Which LLM is good for what?
Yeshwanth Reddy K
9 replies
I mean LLMs although said to be generalised. Some of them are better at few types of workloads. For suppose Claude is way better than any other models when it comes to code. So, I'd like to know what you think about different LLMs.
What do you think about different llms? Do they outperform each other in certain tasks, if yes mention which llms are good for what.
Replies
Abhra Ch.@abhra_ch
@owlmaniac Great question! I totally agree that different LLMs have their own strengths. For example, GPT-4 is excellent at generating creative and detailed responses, while Claude shines in code-related tasks. Then there's LLaMA which is well-optimized for research purposes. Each LLM indeed outperforms others in certain niches! 😄 What about you? Have you noticed any specific strengths or weaknesses in the LLMs you've used? 😊
Share
@abhra_ch I agree with that. I think Claude is for coding I'm constantly using it to generate skeletons. GPT-4 I'm not sure. I think Gemini is good for creative things. Just my observation could be wrong.
I'm sticking with Claude for most part and only think of Gemini and GPT-4 for content creation and human like responses.
Waywise
@owlmaniac I tend to use a mix of GPT-4 for stuff that needs to be a bit more... accurate let's say, whereas I prefer the writing style of Llama 3 which just smells a bit less like it's been written by an LLM. I feel like most people can tell when something has been written by ChatGPT now a mile off!
I don't use LLMs for varied purposes. Just for research & creative writing. I think perplexity is good for factual research & GPT 4 (not 4o) is better for creative writing.
Idk which LLM notion uses- but their "ASK AI" function is also pretty useful when it comes to rewriting copy.
Great discussion! I agree that different LLMs excel in various tasks. For instance, GPT-4 is fantastic for generating detailed and creative responses, while Claude is impressive for coding tasks. LLaMA is often preferred for research purposes due to its optimization.
Have you guys tried using any specific LLM for data analysis or technical writing? If so, which one worked best for you? @owlmaniac @abhra_ch @shearer @my3_murthy
@owlmaniac @abhra_ch @shearer @ashikhameed I don't deal with technical writing. I used to do data analysis back in the day but haven't done it recently.
Navigating through different LLMs is like finding the right emoji - each has its specialty! From GPT-3 acing natural language tasks to Codex ruling the coding realm, these models definitely have their superpowers!