More importantly: for what?

Each model, the tooling you used and even what prompts you use for what model, impacts a lot of the quality of responses you get from the models.