{"id":142115,"date":"2025-08-13T10:15:45","date_gmt":"2025-08-13T10:15:45","guid":{"rendered":"https:\/\/www.europesays.com\/us\/142115\/"},"modified":"2025-08-13T10:15:45","modified_gmt":"2025-08-13T10:15:45","slug":"gpt-5s-model-router-ignited-a-user-backlash-against-openai-but-it-might-be-the-future-of-ai","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/us\/142115\/","title":{"rendered":"GPT-5\u2019s model router ignited a user backlash against OpenAI\u2014but it might be the future of AI"},"content":{"rendered":"

OpenAI\u2019s GPT-5 announcement last week<\/a> was meant to be a triumph\u2014proof that the company was still the undisputed leader in AI\u2014until it wasn\u2019t. Over the weekend, a groundswell of pushback from customers turned the rollout into more than a PR firestorm: It became a product and trust crisis. Users lamented the loss of their favorite models, which had doubled as therapists, friends, and romantic partners. Developers complained of degraded performance. Industry critic Gary Marcus predictably called GPT-5 \u201coverdue, overhyped, and underwhelming.\u201d<\/p>\n

The culprit, many argued, was hiding in plain sight: a new real-time model \u201crouter\u201d that automatically decides which one of GPT-5\u2019s several variants to spin up for every job. Many users assumed GPT-5 <\/a>was a single model trained from scratch; in reality, it\u2019s a network of models\u2014some weaker and cheaper, others stronger and more expensive\u2014stitched together. Experts say that approach could be the future of AI as large language models advance and become more resource-intensive. But in GPT-5\u2019s debut, OpenAI demonstrated some of the inherent challenges in the approach and learned some important lessons about how user expectations are evolving in the AI era.<\/p>\n

For all the benefits promised by model routing, many users of GPT-5 bristled at what they perceived as a lack of control. Some even suggested OpenAI might purposefully be trying to pull the wool over their eyes.\u00a0\u00a0<\/p>\n

In response to the GPT-5 uproar, OpenAI moved quickly to bring back the main earlier model, GPT-4o, for pro users. It also said it fixed buggy routing, increased usage limits, and promised continual updates to regain user trust and stability.<\/p>\n

Anand Chowdhary, cofounder of AI sales platform FirstQuadrant, summed the situation up bluntly: \u201cWhen routing hits, it feels like magic. When it whiffs, it feels broken.\u201d<\/p>\n

The promise and inconsistency of model routing<\/strong><\/p>\n

Jiaxuan You, an assistant professor of computer science at the University of Illinois Urbana-Champaign, told Fortune his lab has studied both the promise\u2014and the inconsistency\u2014of model routing. In GPT-5\u2019s case, he said, he believes (though he can\u2019t confirm) that the model router sometimes sends parts of the same query to different models. A cheaper, faster model might give one answer while a slower, reasoning-focused model gives another, and when the system stitches those responses together, subtle contradictions slip through.\u00a0<\/p>\n

The model routing idea is intuitive, he explained, but \u201cmaking it really work is very nontrivial.\u201d Perfecting a router, he added, can be as challenging as building Amazon-grade recommendation systems, which take years and many domain experts to refine. \u201cGPT-5 is supposed to be built with maybe orders of magnitude more resources,\u201d he explained, pointing out that even if the router picks a smaller model, it shouldn\u2019t produce inconsistent answers.<\/p>\n

Still, You believes routing is here to stay. \u201cThe community also believes model routing is promising,\u201d he said, pointing to both technical and economic reasons. Technically, single-model performance appears to be hitting a plateau: You pointed to the commonly cited scaling laws, which says when we have more data and compute, the model gets better. \u201cBut we all know that the model wouldn\u2019t get infinitely better,\u201d he said. \u201cOver the past year, we have all witnessed that the capacity of a single model is actually saturating.\u201d\u00a0<\/p>\n

Economically, routing lets AI providers keep using older models rather than discarding them when a new one launches. Current events require frequent updates, but static facts remain accurate for years. Directing certain queries to older models avoids wasting the enormous time, compute, and money already spent on training them.<\/p>\n

There are hard physical limits, too. GPU memory has become a bottleneck for training ever-larger models, and chip technology is approaching the maximum memory that can be packed onto a single die. In practice, You explained, physical limits mean the next model can\u2019t be 10 times bigger.\u00a0<\/p>\n

An older idea that is now being hyped<\/strong><\/p>\n

William Falcon, founder and CEO of AI platform Lightning AI, points out that the idea of using an ensemble of models is not new\u2014it has been around since around 2018\u2014and since OpenAI\u2019s models are a black box, we don\u2019t know that GPT-4 did not also use a model routing system.\u00a0<\/p>\n

\u201cI think maybe they\u2019re being more explicit about it now, potentially,\u201d he said. Either way, the GPT-5 launch was heavily hyped up\u2014including the model routing system. The blog post introducing the model<\/a> called it the \u201csmartest, fastest, and most useful model yet, with thinking built in.\u201d In the official ChatGPT blog post, OpenAI confirmed that GPT\u20115 within ChatGPT runs on a system of models coordinated by a behind-the-scenes router that switches to deeper reasoning when needed. The GPT\u20115 System Card went further, clearly outlining multiple model variants\u2014gpt\u20115\u2011main, gpt\u20115\u2011main\u2011mini for speed, and gpt\u20115\u2011thinking, gpt\u20115\u2011thinking\u2011mini, plus a thinking\u2011pro version\u2014and explains how the unified system automatically routes between them.<\/p>\n

In a press pre-briefing, OpenAI CEO Sam Altman touted the model router as a way to tackle what had been a hard-to-decipher list of models to choose from. Altman called the previous model picker interface a \u201cvery confusing mess.\u201d<\/p>\n

But Falcon said the core problem was that GPT-5 simply didn\u2019t feel like a leap. \u201cGPT-1 to 2 to 3 to 4\u2014each time was a massive jump. Four to five was not noticeably better. That\u2019s what people are upset about.\u201d<\/p>\n

Will multiple models add up to AGI?\u00a0<\/strong><\/p>\n

The debate over model routing led some to call out the ongoing hype over the possibility of artificial general intelligence, or AGI, being developed soon. OpenAI officially defines AGI as \u201chighly autonomous systems that outperform humans at most economically valuable work,\u201d but Altman notably said last week<\/a> that it is \u201cnot a super useful term.\u201d<\/p>\n

\u201cWhat about the promised AGI?\u201d wrote Aiden Chaoyang He, an AI researcher and cofounder of TensorOpera, on X<\/a>, criticizing the GPT-5 rollout. \u201cEven a powerful company like OpenAI lacks the ability to train a super-large model, forcing them to resort to the Real-time Model Router.\u201d\u00a0<\/p>\n

Robert Nishihara, co-founder of AI production platform Anyscale, says scaling is still progressing in AI, but the idea of one all-powerful AI model remains elusive. \u201cIt\u2019s hard to build one model that is the best at everything,\u201d he said. That\u2019s why GPT-5 currently runs on a network of models linked by a router, not a single monolith.<\/p>\n

OpenAI has said it hopes to unify these into one model in the future, but Nishihara points out that hybrid systems have real advantages: You can upgrade one piece at a time without disrupting the rest, and you get most of the benefits without the cost and complexity of retraining an entire giant model. As a result, Nishihara thinks routing will stick around.\u00a0<\/p>\n

Aiden Chaoyang He agrees. In theory, scaling laws still hold\u2014more data and compute make models better\u2014but in practice, he believes development will \u201cspiral\u201d between two approaches: routing specialized models together, then trying to consolidate them into one. The deciding factors will be engineering costs, compute and energy limits, and business pressures.<\/p>\n

The hyped-up AGI narrative may need to adjust, too. \u201cIf anyone does anything that\u2019s close to AGI, I don\u2019t know if it\u2019ll literally be one set of weights doing it,\u201d Falcon said, referring to the \u201cbrains\u201d behind LLMs. \u201cIf it\u2019s a collection of models that feels like AGI, that\u2019s fine. No one\u2019s a purist here.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"OpenAI\u2019s GPT-5 announcement last week was meant to be a triumph\u2014proof that the company was still the undisputed…\n","protected":false},"author":3,"featured_media":142116,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[738,64,302,305,923,67,132,68],"class_list":{"0":"post-142115","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-business","8":"tag-artificial-intelligence","9":"tag-business","10":"tag-chatgpt","11":"tag-openai","12":"tag-sam-altman","13":"tag-united-states","14":"tag-unitedstates","15":"tag-us"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@us\/115020937587418392","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/posts\/142115","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/comments?post=142115"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/posts\/142115\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/media\/142116"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/media?parent=142115"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/categories?post=142115"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/us\/wp-json\/wp\/v2\/tags?post=142115"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}