# How to do simple AI text analysis? LLM, fine-tuning, BERT, or embeddings?

Using short-text sentiment, I compare quality, cost, latency, and practicality across four approaches.

## Key points
- Four options side by side
- Results: accuracy, cost, and speed
- Large language model without training
- Fine-tuning a large language model
- Fine-tuned transformer encoder (BERT)
- Embeddings and logistic regression
- My practical recommendation