Introduction and Hype Around GPT-5
At least that is what the hype men and women at OpenAI would like you to think. It is not working on everybody though. That easy benchmark was nothing but a myth. And actually, GPT5 ranks fifth place actually. GPT5 could not surpass Grock on the ARC AGI benchmark, a significant benchmark that has been very conveniently omitted in yesterday announcement. It did not satisfy the betting markets either and OpenAI has now lost its position as the favorite to produce the best model of 2025. The most appalling thing though, several people discovered various issues using OpenAI own charts and this should not be the case when you have the PhD level intelligence in your pocket.
Today, we are going to discover whether GPT5 is yet another over hyped incremental upgrade or a game changer, which is taking us a little further in the fake direction of artificial super intelligence. It is the 8 th August, 2025 and you are watching the code report.
-
USA
- Added by davieasyo
- $0 per hr
davieasyo
Rated: 4 stars
https://www.youtube.com/watch?v=8tx2viHpgA8
The Evolution of GPT Models
Previously, GPT was improved with size. It was being trained using more data that will have more parameters activated during the production of a text. However the era is basically over.
The speciality with GPT5 is not that it is a bigger and smarter model but it integrates a number of models like fast reasoning, routing etc. so that the right tool can be applied to the right task without the end user having to think about it.
GPT5 feels in a lot of ways like a cost cutting/consolidation move after they had been panic shipping a ton of dumbly named models to justify the $200 pro plan during the past year or so.
Pricing and Comparisons
Talking about price, however, GPT5 costs a mere 10 dollars per million tokens of output. That is a lot compared to Claude Opus 4.1 which will burn your wallet at 75 million output tokens per dollar. Sam Alman claims GBT5 is a world of having five to six PhD level professionals in your pocket. However, among the funniest sections of their announcement are the fact that their benchmark graphics are displayed with a y-axis whose scale does not actually mean anything.
And this mistake can have but two causes. One, they vibe charted with GPT5 and it does not in fact possess PhD level intelligence. Or two, they were attempting to be knowingly deceptive.
Better still, GPT5 is allegedly going to have lower levels of deception but then, somebody or something tried to deceive us regarding the Y-axis in the deception benchmark. That is sort of shameful considering a $500 billion intelligence company.
Coding Capabilities Test
However, the biggest question will be on whether GPT5 would be able to program a coded 5 into a spelt app with runes in case you are a programmer. A lot of models have approached it, but they have not exactly succeeded. I tested out GPT5 and I was shocked that it produced gorgeous looking code that spilled, and again did it very quickly, much more rapidly than any other reasoning model. Initially, I imagined I was collaborating with a PhD level spelt developer but this would soon prove to be a disappointment when I ran the code only to get a 500 error as the UI.
What a pain is that GPT5 got the syntax correct, but attempted to apply a rune in the template, which is prohibited. It is expected that GPT5 should record fewer hallucinations, but it hallucinated on using rules of runes. It redeemed itself though when I questioned it again as to whether or not it knew what was wrong with its code, at which time it calculated it out and corrected it. And the final product is a working application with a very nicel looking UI.
I additionally had it create a game of aircraft-flight simulating using 3JS and it came out really awful. The tall person that was from Cursor told me it was the smartest model that they ever used. As much as I would like to think so though, I do not actually believe that GPT5 will kill me or steal my job in the nearest future. Since it is now being made apparent, that the actual potentcy is the ability to mix these new AI instruments together with the old technologies you already underred and adore.