Each time OpenAI releases a brand new type for ChatGPT, the thrill it generates makes me assume I is also lacking out if I do not transfer. The most recent replace is GPT-5.5 Immediate, and it is the default type for all ChatGPT customers. It is paired with an upgraded reminiscence gadget that displays you which of them previous conversations and stored recollections the type pulled from to form its reaction. That is a actually great function that felt lengthy past due. With the brand new reminiscence transparency and claims that hallucinations were decreased by way of over 50% in positive domain names, it is an alluring replace for individuals who had been already at the fence about ChatGPT.
After spending numerous time with each ChatGPT and Claude, my workflow nonetheless hasn’t budged. I am nonetheless achieving for Claude after I want to do genuine paintings, and ChatGPT as my fallback when my Claude utilization creeps too just about the utmost restrict. I do assume that the brand new replace is a real development for ChatGPT. OpenAI thinks so, too, because the statistics they revealed have it ranked above Claude in each and every class. Regardless of what their stats say, my non-public revel in with the fashions tells a distinct tale.
I cancelled my ChatGPT, Perplexity, and Gemini subscriptions for Claude — and I must have quicker
Want I did this quicker.
The reminiscence improve is not actually a a very powerful function
It is not one thing value converting my workflow over
Introducing a solution to see the reminiscence assets for a reaction was once an effective way to handle a not unusual frustration with ChatGPT. In the past, there wasn’t any solution to reliably decide what information the type was once keeping onto, or why it answered how it did. Now, while you obtain a customized reaction, ChatGPT permits you to see which stored reminiscence or previous dialog it drew from. It additionally offers you keep watch over to delete that context or refine old-fashioned entries. Having perception into what influenced the type’s reaction, and having the ability to blank up that context from the similar menu, offers customers rather extra keep watch over.
That mentioned, reminiscence transparency feels extra like a quality-of-life development quite than a jump in capacity. The replace hasn’t modified what ChatGPT is excellent at, or introduced it nearer to Claude’s degree for the issues that in fact subject to me. After a couple of days on GPT-5.5 Immediate, the reminiscence assets function is the only factor I leave out every time I am going again over to Claude. A greater view into what AI recollects about you is good to have, however it isn’t sufficient of a reason why to totally transfer gear. I most commonly use Tasks anyway, the place the reminiscence assets function turns into beside the point. In Tasks, you are able to outline the entire recollections your self and stay the conversations scoped to that context; there is not any query of what influenced a reaction.
The adaptation is apparent in day by day use
For context intensity and coding, Claude stays awesome
I at all times take the printed benchmark numbers with an enormous grain of salt, as a result of they infrequently translate into effects that I will be able to really feel. There are numerous distinctive techniques to make use of AI. A desk of figures is topic to no matter checking out parameters had been used, which have a tendency to serve a selected narrative (each and every corporate will say their type examined best possible), quite than mirror real-world use. In my revel in, Claude has carried out persistently higher than ChatGPT in two large ways in which I regularly use it: lengthy conversations and coding.
The adaptation in context retention is plain after I give the gear a big codebase to paintings on. When running with ChatGPT, I have had it forget about earlier code iterations as soon as the chat grows massive. Every so often it is going to inexplicably trim lengthy sections, reputedly to abide by way of some invisible restrict on how long its reaction is permitted to be. Claude is not flawless both, however I have had fewer of the similar form of incidents with it. It could possibly regularly care for dozens of huge recordsdata, and can proceed regarding parts of them from deep within the dialog. I have heard the similar anecdotal revel in from different builders, and a few unbiased exams concur with my revel in. However like the opposite benchmarks, it most commonly is determined by who you ask.
For coding, the adaptation is much more pronounced. Claude Code is bundled with the Professional subscription and runs for your native gadget. It could possibly learn your code, write to recordsdata, and run exams with out requiring you to replicate and paste context from side to side between your editor and the chat window. Codex is now bundled with ChatGPT Plus, which is available in on the identical worth, so there is not any value benefit both manner. Alternatively, there is a large distinction in how the 2 agentic gear paintings: Codex runs duties in an remoted cloud sandbox, whilst Claude Code operates at once on your native setting. For my setup, that difference issues greater than any benchmark would.
It is the identical solution each and every unencumber cycle
Since each subscriptions value $20 a month, the only to select boils all the way down to which one suits into your workflow higher. ChatGPT is the extra entire product, since it might probably generate pictures and grasp a voice dialog in the similar consultation. However for coding and lengthy paperwork, Claude remains to be the more potent pick out. If ChatGPT’s unique options do not attraction to you (I for my part do not care about symbol era), then Claude’s strengths nonetheless outweigh what OpenAI is providing.
- OS
-
Home windows, macOS
- Person pricing
-
Unfastened plan to be had; $17/month Professional plan
- Team pricing
-
$100/month in step with particular person for the Max plan
Claude is an AI assistant and LLM evolved by way of Anthropic.
