I’m pretty much just using Claude 4 Sonnett and I know that’s both not optimal from a project perspective as well as from a cost perspective.
Any tips on which models are best for specific types of tasks and when we can choose Auto and expect good results?
I usually use Grok-4 and gemini-2.5-probecause its cache reads are not as high as claude-4. Somehow claude-4 always has a very high cache and everyone runs out of tokens.
Auto is always gpt-4.1 I’m desperate with this model