{"id":2490,"date":"2025-04-21T13:06:17","date_gmt":"2025-04-21T13:06:17","guid":{"rendered":"https:\/\/www.stanventures.com\/news\/?p=2490"},"modified":"2025-10-29T07:25:14","modified_gmt":"2025-10-29T07:25:14","slug":"hallucination-rates-spike-in-openais-o3-o4-mini-models","status":"publish","type":"post","link":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/","title":{"rendered":"Hallucination Rates Spike in OpenAI\u2019s o3 &#038; o4-Mini Models"},"content":{"rendered":"<p>OpenAI\u2019s newly released AI models, o3 and o4-mini, are hallucinating at significantly higher rates than previous versions, despite being designed for advanced reasoning tasks.<\/p>\n<p>Internal tests show o3 fabricates information in 33% of factual queries, while o4-mini gets it wrong nearly half the time. The company says it doesn\u2019t yet know why.<\/p>\n<p>The models launched in OpenAI\u2019s next-generation reasoning suite were designed to deliver smarter, context-aware results. However, they have raised a new reliability concern that worries researchers, developers, and enterprise users alike.\u00a0<\/p>\n<p>As hallucinations surge, experts warn these models may be too unpredictable for real-world deployment, particularly in industries where accuracy is non-negotiable.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-2491\" src=\"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png\" alt=\"Hallucination Rates Spike in OpenAI\u2019s o3 &amp; o4-Mini Models\" width=\"1536\" height=\"1024\" srcset=\"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png 1536w, https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM-300x200.png 300w, https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM-1024x683.png 1024w\" sizes=\"auto, (max-width: 1536px) 100vw, 1536px\" \/><\/p>\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_85 counter-hierarchy ez-toc-counter ez-toc-transparent ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\"><\/p>\n<span class=\"ez-toc-title-toggle\"><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1 ' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#the-paradox-at-the-heart-of-openais-reasoning-models\" >The Paradox at the Heart of OpenAI\u2019s Reasoning Models<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#models-that-fabricate-their-own-process\" >Models That Fabricate Their Own Process<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#openai-%e2%80%9cmore-research-is-needed%e2%80%9d\" >OpenAI: &#8220;More Research Is Needed&#8221;<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#the-real-world-risks-of-confidently-wrong-ai\" >The Real-World Risks of Confidently Wrong AI<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#could-search-integration-be-the-solution\" >Could Search Integration Be the Solution?<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#a-broader-challenge-for-the-ai-industry\" >A Broader Challenge for the AI Industry<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#what-users-should-do-now\" >What Users Should Do Now<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-8\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#key-takeaways\" >Key Takeaways<\/a><\/li><\/ul><\/nav><\/div>\n<h2><span class=\"ez-toc-section\" id=\"the-paradox-at-the-heart-of-openais-reasoning-models\"><\/span><b>The Paradox at the Heart of OpenAI\u2019s Reasoning Models<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The launch of <a href=\"https:\/\/openai.com\/index\/introducing-o3-and-o4-mini\/\">o3 and o4-mini<\/a> was supposed to mark a leap forward in artificial intelligence. Positioned as successors to OpenAI\u2019s earlier reasoning models (like o1 and o3-mini), these systems are designed to handle more complex tasks with better contextual understanding, especially in areas like coding, math, and problem decomposition.<\/p>\n<p>But while they may have improved at executing certain logic-driven functions, their overall factual reliability has taken a hit.<\/p>\n<p>In one of OpenAI\u2019s in-house benchmarks PersonQA, which tests how well a model can recall facts about individuals o3 hallucinated on 33% of questions. That\u2019s more than double the hallucination rate of its predecessors: o1 scored 16% and o3-mini, 14.8%.\u00a0<\/p>\n<p>The o4-mini model did even worse, generating incorrect information in 48% of PersonQA cases.<\/p>\n<p>This is not just a statistical anomaly. It\u2019s a systemic regression in one of the most fundamental areas of AI functionality: truthfulness.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"models-that-fabricate-their-own-process\"><\/span><b>Models That Fabricate Their Own Process<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>It\u2019s not only that o3 and o4-mini are getting facts wrong, they\u2019re inventing entire processes.<\/p>\n<p>Transluce, a nonprofit AI research lab, <a href=\"https:\/\/transluce.org\/investigating-o3-truthfulness\">conducted<\/a> independent tests on o3 and observed the model creating fictional narratives about how it arrived at its answers.\u00a0<\/p>\n<p>In one example, o3 claimed that it had run a piece of code on a 2021 MacBook Pro &#8220;outside of ChatGPT&#8221; and then copied the result into its answer.\u00a0<\/p>\n<p>While o3 does have access to some tools, executing code outside its sandboxed environment simply isn\u2019t possible.<\/p>\n<p>This kind of behavior, fabricating not just the answer but the method, adds another layer of concern.\u00a0<\/p>\n<p>When users can\u2019t even trust how a model says it reached a conclusion, the reliability of any AI-generated output comes into question.<\/p>\n<p>Sarah Schwettmann, co-founder of Transluce, summarized the risk plainly: \u201cO3\u2019s hallucination rate may make it less useful than it otherwise would be.\u201d<\/p>\n<h2><span class=\"ez-toc-section\" id=\"openai-%e2%80%9cmore-research-is-needed%e2%80%9d\"><\/span><b>OpenAI: &#8220;More Research Is Needed&#8221;<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>OpenAI, for its part, isn\u2019t denying the issue. In the <a href=\"https:\/\/cdn.openai.com\/pdf\/2221c875-02dc-4789-800b-e7758f3722c1\/o3-and-o4-mini-system-card.pdf\">technical report<\/a> accompanying the model launches, the company openly noted that \u201cmore research is needed\u201d to understand why hallucinations have improved despite progress in reasoning.<\/p>\n<p>One working theory centers on how these models are trained. The o-series models undergo a form of reinforcement learning that may inadvertently amplify hallucination-prone behaviors, especially as they\u2019re optimized to generate more detailed and confident answers.<\/p>\n<p>Neil Chowdhury, a former OpenAI employee and now a researcher at Transluce, believes the reinforcement learning approach might be backfiring.<\/p>\n<p>\u00a0\u201cOur hypothesis is that the kind of reinforcement learning used for o-series models may amplify issues that are usually mitigated (but not fully erased) by standard post-training pipelines,\u201d he explained in an email to TechCrunch.<\/p>\n<p>In essence, the same processes that make the models better at \u201cthinking\u201d may also make them more inclined to confidently fabricate.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"the-real-world-risks-of-confidently-wrong-ai\"><\/span><b>The Real-World Risks of Confidently Wrong AI<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>For consumers casually asking ChatGPT for fun facts or movie trivia, a few hallucinations may seem harmless. But for businesses and professional users, the stakes are much higher.<\/p>\n<p>Kian Katanforoosh, CEO of upskilling platform Workera and a Stanford adjunct professor, has been testing o3 in real-world coding environments. While he praises the model\u2019s advanced capabilities, he notes a recurring problem: \u201cIt tends to hallucinate broken website links,\u201d providing references that don\u2019t exist or lead to dead ends.<\/p>\n<p>That might seem minor in software development, but in other contexts\u2014like generating legal contracts or patient information\u2014hallucinations can result in serious harm, compliance violations, or costly errors.<\/p>\n<p>Enterprises are increasingly interested in AI-powered automation, but persistent hallucination issues can be a dealbreaker. Reliability and factual consistency are not just technical challenges\u2014they\u2019re business imperatives.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"could-search-integration-be-the-solution\"><\/span><b>Could Search Integration Be the Solution?<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>One possible way to curb hallucinations is to give models access to real-time web search. OpenAI\u2019s GPT-4o, which includes a web browsing capability, achieves <a href=\"https:\/\/openai.com\/index\/new-tools-for-building-agents\/\">90% accuracy<\/a> on another benchmark test called SimpleQA.\u00a0<\/p>\n<p>Unlike reasoning models that must \u201crecall\u201d facts from training data, search-enabled models can reference live information to fact-check themselves in real-time.<\/p>\n<p>But there\u2019s a caveat: incorporating web search opens the door to privacy concerns. Queries may be routed through third-party services, exposing sensitive data to external providers. It also introduces new engineering complexity in balancing retrieval with response fluency.<\/p>\n<p>Still, many experts view this hybrid approach\u2014known as Retrieval-Augmented Generation (<a href=\"https:\/\/www.stanventures.com\/news\/what-is-rag-model-how-google-is-using-it-2214\/\">RAG<\/a>)\u2014as one of the most promising paths forward in reducing hallucinations without sacrificing model sophistication.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"a-broader-challenge-for-the-ai-industry\"><\/span><b>A Broader Challenge for the AI Industry<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>The hallucination dilemma arises at a crucial juncture in AI advancement.<\/p>\n<p>In the last year, the industry has predominantly redirected its attention from expanding model sizes to enhancing reasoning\u2014a strategy designed to improve performance while using fewer computational resources.<\/p>\n<p>Reasoning models like o3 and o4-mini reflect this shift, promising better task execution, even with smaller model sizes. But as the latest results show, this shift may come with unforeseen trade-offs.<\/p>\n<p>If increasing a model\u2019s ability to reason simultaneously decreases its ability to stay tethered to facts, the industry will face a hard choice: pursue logic at the cost of truth, or rethink how we build and train AI altogether.<\/p>\n<p>As OpenAI spokesperson Niko Felix told TechCrunch, \u201cAddressing hallucinations across all our models is an ongoing area of research, and we\u2019re continually working to improve their accuracy and reliability.\u201d<\/p>\n<p>But for now, the problem remains unsolved, and potentially getting worse.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"what-users-should-do-now\"><\/span><b>What Users Should Do Now<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>While model developers work on long-term fixes, users can take immediate steps to minimize risks:<\/p>\n<ol>\n<li style=\"font-weight: 400;\" aria-level=\"1\"><b>Always fact-check AI outputs<\/b>, especially when using models in legal, medical, or technical environments.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\"><b>Use models with web access<\/b> when real-time accuracy is more important than prompt privacy.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\"><b>Implement layered review systems<\/b>, such as human-in-the-loop oversight, for critical outputs.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\"><b>Favor retrieval-augmented tools<\/b> that combine generative AI with verified databases.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\"><b>Monitor model updates<\/b> closely, as patch releases may improve accuracy and reduce hallucinations over time.<\/li>\n<\/ol>\n<h2><span class=\"ez-toc-section\" id=\"key-takeaways\"><\/span><b>Key Takeaways<\/b><span class=\"ez-toc-section-end\"><\/span><\/h2>\n<ul>\n<li style=\"font-weight: 400;\" aria-level=\"1\">New models aren&#8217;t always more accurate \u2013 despite better reasoning, hallucinations are increasing.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\">Fabricated processes are a growing issue \u2013 o3 and o4-mini often lie about how they &#8220;know&#8221; something.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\">Businesses must tread carefully \u2013 factual errors make these models risky for critical use cases.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\">Search and RAG may help \u2013 integrating live data sources can reduce reliance on flawed memory.<\/li>\n<li style=\"font-weight: 400;\" aria-level=\"1\">Transparency is essential \u2013 users must know what a model can and can\u2019t do to use it responsibly.\n<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>OpenAI\u2019s newly released AI models, o3 and o4-mini, are hallucinating at significantly higher rates than previous versions, despite being designed for advanced reasoning tasks. Internal tests show o3 fabricates information in 33% of factual queries, while o4-mini gets it wrong nearly half the time. The company says it doesn\u2019t yet know why. The models launched [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":2491,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[15],"tags":[],"class_list":["post-2490","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.8 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Hallucination Rates Spike in OpenAI\u2019s o3 &amp; o4-Mini Models - Stan Ventures<\/title>\n<meta name=\"description\" content=\"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Hallucination Rates Spike in OpenAI\u2019s o3 &amp; o4-Mini Models - Stan Ventures\" \/>\n<meta property=\"og:description\" content=\"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/\" \/>\n<meta property=\"og:site_name\" content=\"Stan Ventures\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/StanVentures\/\" \/>\n<meta property=\"article:published_time\" content=\"2025-04-21T13:06:17+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-10-29T07:25:14+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM-1024x683.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"683\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Dileep Thekkethil\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@dthekkethil\" \/>\n<meta name=\"twitter:site\" content=\"@stanventures\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Dileep Thekkethil\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/\"},\"author\":{\"name\":\"Dileep Thekkethil\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#\\\/schema\\\/person\\\/87d00ff18daf9650e7c925ae4bf86efb\"},\"headline\":\"Hallucination Rates Spike in OpenAI\u2019s o3 &#038; o4-Mini Models\",\"datePublished\":\"2025-04-21T13:06:17+00:00\",\"dateModified\":\"2025-10-29T07:25:14+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/\"},\"wordCount\":1181,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2025\\\/04\\\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png\",\"articleSection\":[\"AI\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/\",\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/\",\"name\":\"Hallucination Rates Spike in OpenAI\u2019s o3 & o4-Mini Models - Stan Ventures\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2025\\\/04\\\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png\",\"datePublished\":\"2025-04-21T13:06:17+00:00\",\"dateModified\":\"2025-10-29T07:25:14+00:00\",\"description\":\"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2025\\\/04\\\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png\",\"contentUrl\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2025\\\/04\\\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png\",\"width\":1536,\"height\":1024,\"caption\":\"ChatGPT Image Apr 21 2025 01 03 22 PM\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Hallucination Rates Spike in OpenAI\u2019s o3 &#038; o4-Mini Models\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#website\",\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/\",\"name\":\"Stan Ventures\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#organization\",\"name\":\"Stan Ventures\",\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Stan-Ventures.webp\",\"contentUrl\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/Stan-Ventures.webp\",\"width\":2001,\"height\":801,\"caption\":\"Stan Ventures\"},\"image\":{\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.facebook.com\\\/StanVentures\\\/\",\"https:\\\/\\\/x.com\\\/stanventures\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/#\\\/schema\\\/person\\\/87d00ff18daf9650e7c925ae4bf86efb\",\"name\":\"Dileep Thekkethil\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g\",\"caption\":\"Dileep Thekkethil\"},\"description\":\"Dileep Thekkethil is the Director of Marketing at Stan Ventures, where he applies over 15 years of SEO and digital marketing expertise to drive growth and authority. A former journalist with six years of experience, he combines strategic storytelling with technical know-how to help brands navigate the shift toward AI-driven search and generative engines. Dileep is a strong advocate for Google\u2019s EEAT standards, regularly sharing real-world use cases and scenarios to demystify complex marketing trends. He is an avid gardener of tropical fruits, a motor enthusiast, and a dedicated caretaker of his pair of cockatiels.\",\"sameAs\":[\"https:\\\/\\\/stanventures.com\\\/news\",\"https:\\\/\\\/www.linkedin.com\\\/in\\\/dileep-pradeep-3705aa53\\\/\",\"https:\\\/\\\/x.com\\\/dthekkethil\"],\"url\":\"https:\\\/\\\/www.stanventures.com\\\/news\\\/author\\\/admin_7mxgn8tx\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Hallucination Rates Spike in OpenAI\u2019s o3 & o4-Mini Models - Stan Ventures","description":"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/","og_locale":"en_US","og_type":"article","og_title":"Hallucination Rates Spike in OpenAI\u2019s o3 & o4-Mini Models - Stan Ventures","og_description":"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.","og_url":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/","og_site_name":"Stan Ventures","article_publisher":"https:\/\/www.facebook.com\/StanVentures\/","article_published_time":"2025-04-21T13:06:17+00:00","article_modified_time":"2025-10-29T07:25:14+00:00","og_image":[{"width":1024,"height":683,"url":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM-1024x683.png","type":"image\/png"}],"author":"Dileep Thekkethil","twitter_card":"summary_large_image","twitter_creator":"@dthekkethil","twitter_site":"@stanventures","twitter_misc":{"Written by":"Dileep Thekkethil","Est. reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#article","isPartOf":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/"},"author":{"name":"Dileep Thekkethil","@id":"https:\/\/www.stanventures.com\/news\/#\/schema\/person\/87d00ff18daf9650e7c925ae4bf86efb"},"headline":"Hallucination Rates Spike in OpenAI\u2019s o3 &#038; o4-Mini Models","datePublished":"2025-04-21T13:06:17+00:00","dateModified":"2025-10-29T07:25:14+00:00","mainEntityOfPage":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/"},"wordCount":1181,"commentCount":0,"publisher":{"@id":"https:\/\/www.stanventures.com\/news\/#organization"},"image":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#primaryimage"},"thumbnailUrl":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png","articleSection":["AI"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/","url":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/","name":"Hallucination Rates Spike in OpenAI\u2019s o3 & o4-Mini Models - Stan Ventures","isPartOf":{"@id":"https:\/\/www.stanventures.com\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#primaryimage"},"image":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#primaryimage"},"thumbnailUrl":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png","datePublished":"2025-04-21T13:06:17+00:00","dateModified":"2025-10-29T07:25:14+00:00","description":"OpenAI\u2019s o3 and o4-mini models promise better reasoning, but they hallucinate more. Experts worry this trade-off could undermine AI\u2019s future.","breadcrumb":{"@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#primaryimage","url":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png","contentUrl":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2025\/04\/ChatGPT-Image-Apr-21-2025-01_03_22-PM.png","width":1536,"height":1024,"caption":"ChatGPT Image Apr 21 2025 01 03 22 PM"},{"@type":"BreadcrumbList","@id":"https:\/\/www.stanventures.com\/news\/hallucination-rates-spike-in-openais-o3-o4-mini-models-2490\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.stanventures.com\/news\/"},{"@type":"ListItem","position":2,"name":"Hallucination Rates Spike in OpenAI\u2019s o3 &#038; o4-Mini Models"}]},{"@type":"WebSite","@id":"https:\/\/www.stanventures.com\/news\/#website","url":"https:\/\/www.stanventures.com\/news\/","name":"Stan Ventures","description":"","publisher":{"@id":"https:\/\/www.stanventures.com\/news\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.stanventures.com\/news\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/www.stanventures.com\/news\/#organization","name":"Stan Ventures","url":"https:\/\/www.stanventures.com\/news\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.stanventures.com\/news\/#\/schema\/logo\/image\/","url":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2024\/06\/Stan-Ventures.webp","contentUrl":"https:\/\/www.stanventures.com\/news\/wp-content\/uploads\/2024\/06\/Stan-Ventures.webp","width":2001,"height":801,"caption":"Stan Ventures"},"image":{"@id":"https:\/\/www.stanventures.com\/news\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/StanVentures\/","https:\/\/x.com\/stanventures"]},{"@type":"Person","@id":"https:\/\/www.stanventures.com\/news\/#\/schema\/person\/87d00ff18daf9650e7c925ae4bf86efb","name":"Dileep Thekkethil","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/911bd385b9da54d4a69f19f536a6419e576244371bd6e7d96f06c583dd402fa9?s=96&d=mm&r=g","caption":"Dileep Thekkethil"},"description":"Dileep Thekkethil is the Director of Marketing at Stan Ventures, where he applies over 15 years of SEO and digital marketing expertise to drive growth and authority. A former journalist with six years of experience, he combines strategic storytelling with technical know-how to help brands navigate the shift toward AI-driven search and generative engines. Dileep is a strong advocate for Google\u2019s EEAT standards, regularly sharing real-world use cases and scenarios to demystify complex marketing trends. He is an avid gardener of tropical fruits, a motor enthusiast, and a dedicated caretaker of his pair of cockatiels.","sameAs":["https:\/\/stanventures.com\/news","https:\/\/www.linkedin.com\/in\/dileep-pradeep-3705aa53\/","https:\/\/x.com\/dthekkethil"],"url":"https:\/\/www.stanventures.com\/news\/author\/admin_7mxgn8tx\/"}]}},"_links":{"self":[{"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/posts\/2490","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/comments?post=2490"}],"version-history":[{"count":1,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/posts\/2490\/revisions"}],"predecessor-version":[{"id":5139,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/posts\/2490\/revisions\/5139"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/media\/2491"}],"wp:attachment":[{"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/media?parent=2490"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/categories?post=2490"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.stanventures.com\/news\/wp-json\/wp\/v2\/tags?post=2490"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}