Post
Conversation
![](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="73" height="73"><rect fill-opacity="0"/></svg>)
![David Watson 🥑](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="96" height="96"><rect fill-opacity="0"/></svg>)
Post your reply
Pretty interesting discussion in the thread:
I feel like everyone is sleeping on Gemini. Maybe because it doesn’t have the same flashiness (no pun intended) because it’s from a big boring corp, but I wouldn’t be surprised if it takes a solid lead in the upcoming months.
After GPT-5 launch, the chance of OpenAI winning the best model was decreased by 20% 
You knew things weren’t good when they invited influencers instead of pros to evaluate it
ty. I'm so fucking tired of people comparing AI to Nukes, Manhattan project, the end all be all tech, and now the death star.
All it does is scare people, bring regulatory heat, and then force the labs to bitch they are getting regulations they brought onto themselves.
Trying to work with GPT5 on Cursor felt more or less the same:
1) Overthinks for even smaller tasks.
2) Still fu**ks up the task.
x25
Trying this only because it's free on .
Claude Sonnet 4 somehow knows what to focus on instantly as I ask it.
Reasoning kinda
>GPT-5 is about lowering costs for OpenAI
Yes, here’s why 
Quote
Cloud Ventures | Cloud Deals & FinOps 
@cloudventures_
Here’s why OpenAI did this
And yes, it comes down to optimizing their cloud costs
Maintaining 10 different models requires multiple GPU fleets to have the capacity so serve all model types and users
Whereas going all-in on GPT-5 allows them to scale up to one large x.com/levelsio/statu…
Show more
Well it is Reddit. They haven’t liked anything for at least several years
GPT 5 has overall gotten better. People are just impatient and jumping the gun after UX surprises.
I’m really enjoying the update. Haven’t tried Claude yet, but with GPT-5 and the new features, there’s nothing pulling me away. The writing is amazing, it handles math, coding, and design well, and it fits smoothly into my workflows.
Seems many people are disappointed, but I’m
sad to see OpenAI cutting model options and context window for Plus users. Maybe the time to migrate to Claude, Grok or Gemini
Lowering costs helps all AI products improve. They’re more affordable and able to increase context size.
Personally I think the hate on OpenAI is just posturing for likes/attention.
To be fair, competition is fiercer so their products aren’t as amazing anymore.
Still,
Glad that X and Reddit audience doesn't act like paid youtubers who were glazing gpt5 all day.
gpt-5 pro is amazing AF… still need to compare to o3 and o3 pro but initial impression is that it’s an amazing model — better than o3
openai is becoming the yahoo of foundation ai providers
too much talent bleed, too much competition
Some people have criticized GPT-5 after its release, but I have noticed that other experts, such as quantum physicists, have been stunned by its capabilities. It is important to judge a model with a broader perspective, looking beyond our own examples, because GPT-5 shows
would have been more impressed with GPT-5 if hadn't used claude opus 4.1 just a day before.
Agreed! GPT-5 is alright and not much there. Claude keeps being my number 1 and I have been using the combo on Windsurf by using GPT-5 (low reasoning) for planning with Claude Sonnet 4 to execute the plan. Overall, it's an mild improvement for OpenAI.
Boggles my mind how there is such a huge divide on this. People that seem to have early access like and are raving. Then there are post like this. Is it totally subjective and only up to the vibes?
Used 5 extensively today. Have been a daily user for a year. 5 blows. Completely awful.
pretty much a nothing burger wrapped in a fart. Claude Code still reigns supreme.
It’s beaten hollow Claude in cursor for my code base. The thing with models is u need to choose what works best for you. I am pretty sure for another project Claude will do better.
Also, that’s a good litmus test on people who use models and who actually use models
I was banned from that sub for saying that Grok had some features that were better.
Reddit is over.
I have had three instances of unsafe Rust in a storage engine I’ve been building for, well, a while. I’ve tried to get them refactored out manually; tried with Claude Code, Gemini, GPT o3/o1 Pro. I’ve read the Nomicon end to fucking end and I just couldn’t get it done.
GPT5
Absolutely. Routers are an optimization choice more than a UX choice here
OpenAI is now mainly a product company.
They are trying to simplify the product via orchestration and aiming to make prompts go to the most cost effective model.
I haven't used it enough yet but so far it seems fine. The main issue power users are hitting is the orchestration
Respect for acknowledging the public feedback and jumping on it. Having said that If GPT-5 was only good and no slop, OpenAI wouldn’t have to do a immediate AMA and take corrective actions to roll back actions and make changes.
Yeah the death star thing was really overhyping. Was really expecting more development from these companies this year -- things have plateaued a bit.
claude still reigns supreme for me. GPT5 has been terrible for everything i've asked it to do.
I’ve been underwhelmed. It seems like a branding effort where they combined 4o and o3 into a single interface.
Except 4o seems worse.
A little overstated in my opinion but I do agree it was a bit underwhelming and feels somewhat cost motivated. Plus I kinda miss being able to choose models
Agree totally. While the hype around AGI should continue, the GPT-5 update feels more like a consolidation aimed at reducing cognitive load for regular users.
Before the launch, I often found 4o better than O3, and as the release approached, 4o even seemed smarter, which is why
Low taste users are the majority, got it
Quote
gary IH fung
@garyfung
Yup. For complainers that gpt5 sucks: prompting skill issue
All the more reason to use functional equivalent of Plan mode. For serious work, coding (or otherwise), chat or manually iterate on writing a detailed plan before executing the coding x.com/cline/status/1…
Tossing GPT-5 under the bus and hyping up Claude’s reign is a bold swing!
I’m itching to see the data prove that throne’s legit
He’s not wrong and they also cut the context window for plus from 200k to 32kz.
I mainly use mine for personal research and I haven't noticed any real improvement. In fact I had to restate directions multiple times for GPT5 to do what I wanted it to. I haven't had that problem much in the past.
wild how the #1 takeaway from GPT-5 launch week is basically: Claude supremacy, vibes check, and cost-cutting confirmed.
I think all that matters now is Claude can actually code decently well in a bottle of tasks.
If you’re hand holding it somewhat it still takes time but gets things done immaculately. Like not vibe code but more pointedly tell it what lines you want edited and files to think
Kind of agree, except the Death Star is for other AI labs who are going to need a hard haircut to compete with gpt5 free for all and the new API pricing.
On a podcast a few months back he said about for O3 users " I feel really bad for the people using the current model , if they had what I am using they would be so blown away it doesn't even compare"
Seems like o3 was peak for them. Still think agent mode/operator for the right
I was disappointed too, until I tried one-shotting a good-looking mini-game with GPT-5 in Python. Feel the upgrade 
The paranoia is already improving! A great sign that the sycophancy damage is already healing.
I feel like claude is much better compared to GPT-5 for coding related to tasks
100% right on. This is very much my experience. I'm honestly super disappointed.
Someone had to say this.
Especially after knowing about the rate limits on using 5 (thinking) or even 5 even being a plus user and forced change of models. 
I tried gpt5 for a coding task of moderate difficulty and had to switch back to claude 4 to get both faster and better quality code edits.
Mass-market always makes design difficult.
Claude is good but can barely keep up, they have to limit its use.
Def right about the hype, I saw the Death Star and was like "look, the bubble!" lol
ChatGPT’s status as the default choice for beginners comes with a significant test-time cost.
With switching costs so low, this seems less like a competitive moat and more like a cash-burn handicap.
All the doom and gloom talk from Sam is purely for marketing.
He took that play right from Elons book.
Mortals, Congratulate yourselves. AGI is cancelled and your pathetic jobs are safe.
ChatGPT is getting worse with each update. I’ve had the premium for years; it has not improved - each update scraps any real genius and replaces it with a calculator.
Claude 3.7 for coding so far….it ain’t the flyest ride, but it’s reliable.
So real ..Claude is still ahead and I for one am happy about it. .
I’ve always said it, claude is far more better thay ChatGPT
For me, it still comes back to one thing: trust.
Without it, the rest is just a demo.
LLMs are on plateau, improvements are mostly from adding MCP and other tools to it, like searching tweets or storing some data in RAG.
And they still can't reason, only imitation.
OpenAI is using Reddit data for learning, grave mistake.
If GPT-5 were an iPhone release, based on the hype & improvements, which would it be - iPhone 4, iPhone X, or iPhone 14?
it isnt a bad thing imo
lowering cost means far efficient systems, and overall less computation needed for the same results
its useless for even simple tasks
Quote
Paul Tan
@paultantk
this is so unhelpful @OpenAI @sama
Popularity is interesting. What drives it in online communities? Is it quality, timing, or something else?
