Hi guys! I’m using the chat GPT to assist me with programming. Since the o1 release I was extremely satisfied about coding powers of gpt with my Plus subscription BUT. Since yesterday I’ve noticed that o1 and gpt-4o started halucinate really hard in code writing. Almost all of 20 coding requests it’s used to nail perfectly before now contain mistakes and non-existent methods and packages within requested tech stack and so on, for example it started using React hooks in Vue code considering that I’ve never mentioned React in conversations before. Or it stopped to provide a valid Bun server setup as it has forgotten what it is, at the time that few days before I’ve build mostly entire Bun project using gpt. I feel like I’m using gpt3-5turbo because of response quality.
Does some face issues with response quality too?
Same here. Now GPT4 o1 Plus is worse than GPT4 mini o1 and even GPT 3.5
I bet 2 possibility going on.
-
OpenAi is testing the water by reducing computing resources so that users get stress out and miss the last GPT4 o1 preview’s excellancy. Now they put out the 200$ price tag PRO mode that features infinate number of prompting. Definately they are going to raise the PLUS plan price and get back the resources but still would be limited or keep the poor Plus plan hoping to attract PLUS users to PRO mode plan. Anyway both are taking more money from user’s pocket.
-
OpenAi is temporarilly reduced computing power due to the GPU demanding service
like Sora and o3. In my country Sora was inaccessible for a while for overcrowded user. And I read an wallstreet journal article that it takes more than 0.5 billion dollars to train the o3 service for 6 monthes. That why o3 launch is being delayed. Bottom line. It’s about their budget and the current dumb GPT4 o1 is part of that. However this is better because it’s temporary decline.
I was a stock broker but transfered to AI investment team. And there I do code job in the office even though I never majored in computer science. I’d been really satistied with GPT4 o1 preview. Sometime my project was way better than my coworker who is proffesional computer programmer. But now I’m looking for the competition and found Gemini 2.0 advanced 1206 as good as GPT4 o1 preview. I successfully wrote codes today with Gemini 2.0 Advanced which is 1 month free. To be honest still GPT4 o1 preview slightly superior to Gemini 2.0 advanced 1206. But I’m considering otherwise according to OpenAi’s next move.