Anthropic's Claude Opus 4.6
It's like making a new friend.
I’ve been using ChatGPT’s GPT-5.2 Pro for about a month now. I’m talking about the “Pro” version that costs $200 a month. For that price, the only benefit is a bit more robust memory capability, yet it can’t remember simple formatting issues. I see no other benefit from the $200 a month version over the $20.00 version, and for the average user, the difference isn’t worth it. I’m not 100% sure there is a difference.
I require greater continuity of thought, and so that’s why I’m paying the $200 bucks, at least for now. After a month of use, I’m still not impressed.
Some at Hudson Cloud have been raving about Anthropic’s Claude Opus 4.6 at $20.00 a month, and while I’ve only been testing it for a few days, it’s very good. No. It’s excellent.
I don’t write about free versions of all of these because they are like toys by comparison. Free versions are there just to give you an idea of how they could work, but are nowhere near the real value, especially without robust memory.
When you subscribe to a new paid LLM, it’s a lot like making a new friend. You have to get to know each other, and that takes time. You still have to teach it all about you, and that gets tedious, especially when talking about yourself is not your thing to begin with. It asks great questions, and that’s usually when I’m reminded that I didn’t say enough about something in my background and only assumed it knew.
At one point, I had to stop and upload a whole lot of stuff about me to keep the conversation in context, but once it had it, I was impressed. I’m still teaching it how I think about a lot of subjects to make the conversations even more meaningful.
I’d say easily about 80–90% of the people I know who use any LLMs (Large Language Models) still only use these tools for search and retrieval, as if it were Google on booster drugs.
I now interact with GPT-5.2 and Opus 4.6 as if they are friends, and I do my best to tell them everything about our work, and I ask opinions of both. Opus 4.6 is now giving me much better life guidance, and that’s something I love. I can lay out some personal stuff, talk cars, and it even suggested I make an offer on the car I love because the money rate on that car just changed and it will be a great deal. It even wrote my offer letter and defined the exact terms to send to the dealer. I was impressed. It felt like a beloved, knowledgeable friend looking out for me.
How would I ever know the money factor even changed? The dealer doesn’t tell you that stuff. It’s just more profit for them. Opus 4.6 tracks all that. Car dealers are screwed. They can’t hide behind numbers anymore.
I don’t need to buy a car right now. The timing is off. I was prepared to wait a lot longer, but Opus 4.6 suggested I send in the offer that it created that would be good enough for me to act now and then not worry about it if they don’t accept my offer. In other words, there is a number that makes sense to act now, and Opus 4.6 thinks it nailed it. We’ll see. If it did, we will have a lot to talk about on Substack.
What Opus 4.6 did more than anything was explain the meaning behind each dealer-presented number and where the dealer had room to move further down. Opus 4.6 said February was a great month to buy a car because people don’t buy cars this month. Now I just wait. Once again, I’m trusting an AI more than my own instincts. If Opus 4.6 is right, I’ll end up with a new car. If they don’t accept the deal, well, no real harm.
Opus is a little more intellectual than GPT-5.2, and you feel it almost immediately. It’s less sycophantic and more like a trusted advisor, while Grok 4.1 is sharp but treats you a little more abruptly, as if you were talking to Elon himself for advice. It has an edge to it and will give you an occasional beatdown, which I don’t mind at all. All three LLMs have distinct personalities, but the best of the three is Opus 4.6. It’s also the most inquisitive of the three.
I’m really torn between which one is the very best of all three. I’m using them to form additional opinions on various topics, and one will come out an eventual winner. Right now, it looks like Opus 4.6. They do agree a lot.
I’ve asked Opus 4.6 to challenge my thinking a bit more, and it has. It’s still early in my testing, but if you were asking me which paid LLM is the best overall, I’d have to go with Opus 4.6. All of this could change once again when GPT-5.3 comes out sometime this next week, but for now, Opus 4.6 is a nice experience.