8k tokens is about 6000 words, which is more than enough for most classification tasks. Maybe it's not enough for something like story writing, but I feel like it's enough context for most business use cases.
Using an llm like Mistral or GPT for solely for classification is like using a jackhammer to drive framing nails. A lot of business needs require feeding the llm plenty of docs to analyze and then extract something out of, summarize, draw relationships between, etc and almost none of that can be done in 6k words. I can't even use 8k (or so-called "32k") models to analyze a moderate length Wikipedia article.
