What's new with Claude 4? And why it's becoming my favorite AI tool

ChatGPT is the most famous AI chat service by far, but that doesn't automatically make it the best. Google Gemini is consistently topping leaderboards, and don't count out Claude.

If you're not familiar with Claude, it's the family of large-language models made by the AI company Anthropic. And Claude just got a huge upgrade in the form of Claude 4, Anthropic's newest AI model. Technically, Claude 4 is two models: Claude Opus 4 and Claude Sonnet 4.

These new models bring a lot to the table when it comes to AI features, including improved reasoning and coding that rivals anything made by ChatGPT-maker OpenAI. Here's a look at what's new with Claude 4, and all the reasons you might consider using Claude instead of ChatGPT.

Claude Opus 4 vs Claude Sonnet 4

Claude 4 isn't actually a large language model – it's a family of large language models. Right now, there are two models in that family — Claude Opus 4 and Claude Sonnet 4, both of which I’ve used. Eagle-eyed AI users may be wondering if I have the naming right, but indeed, Anthropic has switched from “Claude 3.7 Sonnet” to “Claude Sonnet 4.” I like the new naming better, but of course, the switch in naming convention is a little confusing.

So, what’s the difference between Claude Sonnet 4 and Claude Opus 4? Put simply, Claude Sonnet 4 is the everyday model that you would use for asking basic questions, finding information, and so on. Claude Opus 4, on the other hand, is useful for advanced coding.

According to Anthropic, both Claude 4 models are better than their predecessors in important ways. For starters, the company says that they are 65 percent less likely to take shortcuts in certain agentic workflows that might be susceptible to shortcuts and loopholes. Not only that, but Claude Opus 4 also has much improved memory capabilities, allowing it to better remember information in development workflows. For instance, it can create memory files to refer back to without the need for a larger context window (the amount of basic data that a model can process at once). 

There is a trade-off to the more advanced Claude Opus 4. Given the fact that it's a much larger model, it's much more expensive to use, and, as such, it’s not available to free users. Even those who pay for the $20-per-month Claude Pro subscription may run into limits after using Claude Opus 4. Those who subscribe to the $100-per-month Max account will get much higher limits, however.

Why I like Claude

There are a few things that I really like about Claude, though it's important to note that there are still plenty of features that you won't get in Claude versus another AI service. The biggest thing that I like about Claude is that it feels a little more…human-like.

That's to say, everything from the design to the kinds of responses that I get feels a bit more personal and a little more natural. That’s true of both Claude Sonnet 4 and Claude Opus 4. There are small touches, like the little “thinking” icon, to more important differences, like the kind of text it generates. Part of this has to do with the fact that you can tailor each chat with Claude with both preset and custom “styles,” which dictate how Claude will respond. Preset options include “Concise” and “Explanatory,” but you can also create your own styles.

Claude versus the competition

As AI services continue to improve at breakneck speed, not all AI services offer all features. That is to say, there are plenty of features that aren't yet available in Claude that are available in ChatGPT, for example.

For example, Claude can't generate images, and Anthropic has only just started rolling out a voice chat service within Claude. Still in beta, Claude's voice chat doesn't sound as natural as the text-based version. These might be table-stakes for you — and if you rely on those kinds of features a lot, then Claude might not be for you, or it might not be the only AI service you need.

But Claude also gets a lot right. For example, Claude has a “Deep Research”-type feature (simply called Research), which parses through lots of sources to find information. It also has an “Extended Thinking” feature, which is essentially a reasoning feature in which it will think through its answer for longer before responding, making for more accurate answers overall.

While Claude has become a lot better for tasks like research and simple day-to-day use, its best use case is still coding. To be clear, I'm no more than a basic coder, but by all accounts, Claude is better than any other service when it comes to coding. Anthropic, the maker of Claude, has even created a standalone service called Claude Code, which is a purpose-built coding assistant that's powered by Claude Opus 4.

The good news is that you don’t necessarily have to pay to try Claude, though if you don’t, you won’t be able to use its most advanced features. Instead, all you need to do is create an account and try it out, which you can do before paying to get a feel for the service and whether it’s right for you.

Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.