{"id":4506,"date":"2025-08-14T19:50:45","date_gmt":"2025-08-14T19:50:45","guid":{"rendered":"https:\/\/musictechohio.online\/site\/theory-why-gpt-5-sucks\/"},"modified":"2025-08-14T19:50:45","modified_gmt":"2025-08-14T19:50:45","slug":"theory-why-gpt-5-sucks","status":"publish","type":"post","link":"https:\/\/musictechohio.online\/site\/theory-why-gpt-5-sucks\/","title":{"rendered":"There’s a Compelling Theory Why GPT-5 Sucks so Much"},"content":{"rendered":"
\n
\"The<\/div>\n

The launch<\/a> of OpenAI’s long-awaited GPT-5 model last week marked an ignominious chapter for the company. While it boasted modest performance upgrades on paper, the actual experience of using it left many fans <\/span>sorely disappointed<\/span><\/a>.\u00a0<\/span><\/p>\n

Common criticisms were that its answers were too short and its writing was noticeably worse and devoid of personality. On top of that, many noted that despite its “PhD level” intelligence, it still made <\/span>dumb errors<\/span><\/a> like insisting there are three Bs in the word “blueberry.”\u00a0<\/span><\/p>\n

It was such a letdown that fans demanded that OpenAI give them back access to the previous model GPT-4o, which the company boldly removed with the launch of GPT-5. And stunningly, OpenAI capitulated to fans’ demands<\/a>.<\/span><\/p>\n

In sum, it did not feel like a “significant step along the path to AGI,” as CEO Sam Altman bragged it was.<\/span><\/p>\n

There may be an interesting reason behind why the ChatGPT upgrade feels so lackluster. As <\/span>The Register <\/span><\/i>speculates<\/span><\/a>, it’s likely because GPT-5 is really meant to be “less of an advancement and more of a way to save compute cost.”<\/span><\/p>\n

It’s a striking theory, highlighting how the AI industry continues to burn through billions of dollars, eclipsing moderate revenues. OpenAI could be looking to pump the brakes to better its chances of turning a profit before the end of the decade, but whether that’s even a possibility at this point remains a point of contention.<\/p>\n

GPT-5 isn’t a single model \u2014 it’s actually a tag team of a lightweight model to handle basic requests, and a beefier one to tackle complex “reasoning” tasks. A separate router model chooses which of the two LLMs should tag in, based on the prompt.<\/span><\/p>\n

As it turned out, the router model sucked at its job, and “broke” on launch day, Altman claimed<\/a>, making GPT-5 seem “way dumber.”\u00a0<\/span><\/p>\n

It’s now <\/span>back online<\/span><\/a> with some purported improvements, but has done little to conciliate fans’ fury at having their choices limited.<\/span><\/p>\n

The important takeaway from this is that deploying this “autoswitcher” strays from OpenAI’s approach in the past, <\/span>The Register <\/span><\/i>notes, which allowed paid users to simply select which model they wanted to use, instead of it being chosen automatically. Keeping all those models online, however, is expensive, reinforcing the idea that the ChatGPT maker is undergoing some belt-tightening by getting rid of them \u2014 though it has since reversed course and restored access to paid users.<\/span><\/p>\n

There are other blatant signs of cost-cutting, like a severe limit of just ten messages per hour for free users. OpenAI is also keeping the model’s “<\/span>context window<\/span><\/a>” \u2014 essentially its memory \u2014 the same as before, limited to 32,000 tokens for Plus users and 128,000 for Pro. Fans, especially those in the Plus tier \u2014 which at $20 is the only affordable tier for most, as Pro will set you back $200 per month \u2014 have been <\/span>begging<\/span><\/a> for<\/span> an upgrade in this area.<\/span><\/p>\n

GPT-5 representing behind-the-scenes pruning at the company certainly been the theory held by many of its fans, with one of the top posts on the r\/ChatGPT subreddit <\/span>averring<\/span><\/a> that GPT-5 is “clearly a cost-saving exercise.”<\/span><\/p>\n

“They removed all their expensive, capable models and replace[d] them with an auto-router that defaults to cost optimisation,” the user wrote. “That sounds bad, so they wrap it up as [GPT-5] and proclaim it’s incredible.”<\/span><\/p>\n

“Feels like cost-saving, not like improvement,” <\/span>surmised<\/span><\/a> another.<\/span><\/p>\n

Unglamorous as it is, cost-cutting at this moment makes sense from OpenAI’s point of view. It’s facing more competition than ever and is <\/span>under increasing pressure<\/span><\/a> to find a way to turn its business model profitable. Its <\/span>anticipated valuation<\/span><\/a> of some $500 billion comes with the implicit expectation that it will figure out how to make money soon. <\/span><\/p>\n

But the company clearly underestimated just how wildly attached fans would become to the quirks of its older models, even if they are nominally inferior \u2014 and that’s not a problem that’s going to go away anytime soon.<\/span><\/p>\n

More on AI: <\/strong>Scientists Are Getting Seriously Worried That We’ve Already Hit Peak AI<\/a><\/em><\/p>\n

The post There’s a Compelling Theory Why GPT-5 Sucks so Much<\/a> appeared first on Futurism<\/a>.<\/p>\n<\/div>\n

<\/div>","protected":false},"excerpt":{"rendered":"

The launch of OpenAI’s long-awaited GPT-5 model last week marked an ignominious chapter for the company. While it boasted modest performance upgrades on paper, the actual experience of using it…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[177,196,183,2935,179],"tags":[],"class_list":["post-4506","post","type-post","status-publish","format-standard","hentry","category-artificial-intelligence","category-chatgpt","category-generative-ai","category-gpt-5","category-openai"],"_links":{"self":[{"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/posts\/4506","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/comments?post=4506"}],"version-history":[{"count":0,"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/posts\/4506\/revisions"}],"wp:attachment":[{"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/media?parent=4506"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/categories?post=4506"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/musictechohio.online\/site\/wp-json\/wp\/v2\/tags?post=4506"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}