Which LLMs are the best value for money?
The cost of LLMs is steadily falling, and the quality is rising.
A rough estimate of the cost of an LLM is the cost per million tokens of input, mostly from LLMPriceCheck. (Typically, inputs are the bigger component of the cost, compared to outputs.)
A rough estimate of the quality of an LLM is the ELO score on the LMSYS Leaderboard. (This is like the chess ELO score, but for LLMs, where people compare 2 LLMs on the same task.)
Some LLMs are "pareto optimal" — no other LLM is better in both cost and quality. These are shown in 🟢 green and are the best to use.
Some LLMs are "pareto suboptimal" — no other LLM is worse in both cost and quality. These are shown in 🔴 red and are the ones to avoid.