After trying GPT-5, the responses definitely sound more natural
I used to use GPT-4o when I needed to draft documents at work, but since GPT-5 came out, I tried the beta.
Overall, its contextual understanding has improved a lot, and it really maintains consistency, especially in long-form writing. It still occasionally stumbles on technical terms in certain fields when writing in Korean, but the overall quality is satisfying.
If you're considering adopting it, I'd suggest calculating the cost-effectiveness first. Choosing the right plan will depend on your monthly usage.
10 answers
Yeah, I've tried it too, and it's really natural now. But it's expensive.
Oh, I didn't know this. I'm still using GPT-4, but I was thinking about switching, so I'll keep this in mind.
Well, I'm not so sure about that. I actually find GPT-5's answers get too long and become distracting. Even when I ask it to keep it short, it feels like it habitually rambles on.
I've felt that slip-up with Korean technical terms too. When I have the legal team work on documents, they get the clause names wrong—yeah... it's not quite perfect yet.
If it's a beta version, we'll just have to wait and see. 4o was said to be good at first too, but it turned weird after updates.
At my previous job, I ran document drafts through GPT and it saved a ton of time. But there are still spots that need a human touch. By the way, using the API is cheaper.
As for cost-effectiveness... our team just decided to stick with 4o. Even switching to 5, the benefit wasn't noticeable.
I'll admit the long-form consistency. I had it write a 20-page report, and the cases of inconsistencies dropped significantly. However, it's not wrong to say that Korean data is a bit lacking.
Got a source? There are a lot of beta review posts that seem like ads.
Haha, I still miss 3.5. Honestly, if this is what I expect from AI, I'm satisfied.