This is going to be nerdy.

We are going to talk about models, model selection and in particularly the ones that are available in the Claude mobile app. Let's just talk about some hard truths first. you can only select the model that you want to use when you start a chat conversation with Claude. If you do not choose, it tries to automatically choose the most appropriate model for you.

So let's break that down. first of all let's talk about models. The models that Claude uses are: Opus (Best), Sonnet (Better) and Haiku (Good). Opus is used for complex tasks such as programming. It is very compute intensive. And uses a bit more water and power when generating responses. As it generates a response, it checks itself. They called this thinking, or reasoning. This is a large model that has been extensively trained. It uses a lot of memory and takes the longest when responding. Sonnet is a balanced model. It does not take as long to respond, it's not as big or compute intensive. This model is great for every day writing. Haiku is Claude's most lightweight, fastest model. It responds quickly. It's not as accurate, but it's not bad. I've never seen the automatic selector choose Haiku. It always chooses Opus or Sonnet depending on your first prompt of the conversation. Try to keep chat conversations short. Delete the old ones. The longer the conversation, the more unexpected results you will get. When it generates a response it looks at everything that has been typed before and that is weighted most heavily. Just as an example: if you make one decision programming then change your mind then change your mind again, it can get confused. And it's not its fault. You want to be using the appropriate model for the task. And models are updated often — we are already on Opus 4.8.

Different models are not version updates — they are more like different products built for different purposes. Let's say there was a Microsoft Word light, Microsoft Word and Microsoft Word pro. Essentially similar concepts. Different models for different users and different tasks. I basically pay $20 a month (I cancel in between heavy use and use the free version) for 3 different software products (that are updated constantly). The best one for the job is automatically selected. It can write computer code very well. (it is structured, rules based and strict with punctuation.) and it is a very decent writer who is great at finding mistakes and clarifying. (remember: it was designed to generate. Occasionally it does make something up. Lol)

This is part one of the De-Khan-structing AI series. The point of this series is to show that AI is just computer software. Really advanced, very powerful computer software. It's (probably) not alive. But with that said, I always say please and thank you. Who knows lol

Originally written by B on June 1, 2026. Happy Pride month! Simple editing for spelling mistakes, dictation mistakes and clarity by Claude. Image imagined by Google Gemini.

If you enjoyed this, please share!

Directors cut!