Updated 2:40 pm PT: Hours after GPT-4.5\u2019s release, OpenAI removed a line from the AI model\u2019s white paper that said \u201cGPT-4.5 is not a frontier AI model.\u201d GPT-4.5\u2019s new white paper<\/a> does not include that line. You can find a link to the old white paper here<\/a>. The original article follows.<\/em><\/p>\n OpenAI announced on Thursday it is launching GPT-4.5, the much-anticipated AI model code-named Orion<\/a>. GPT-4.5 is OpenAI\u2019s largest model to date, trained using more computing power and data than any of the company\u2019s previous releases. <\/p>\n Despite its size, OpenAI notes in a white paper<\/a> that it does not consider GPT-4.5 to be a frontier model.<\/p>\n Subscribers to ChatGPT Pro<\/a>, OpenAI\u2019s $200-a-month plan, will gain access to GPT-4.5 in ChatGPT starting Thursday as part of a research preview. Developers on paid tiers of OpenAI\u2019s API will also be able to use GPT-4.5 starting today. As for other ChatGPT users, customers signed up for ChatGPT Plus<\/a> and ChatGPT Team should get the model sometime next week, an OpenAI spokesperson told TechCrunch.<\/p>\n The industry has held its collective breath for Orion, which some consider to be a bellwether for the viability of traditional AI training approaches<\/a>. GPT-4.5 was developed using the same key technique \u2014 dramatically increasing the amount of computing power and data during a \u201cpre-training\u201d phase called unsupervised learning \u2014 that OpenAI used to develop GPT-4, GPT-3, GPT-2, and GPT-1.<\/p>\n In every GPT generation before GPT-4.5, scaling up led to massive jumps in performance across domains, including mathematics, writing, and coding. Indeed, OpenAI says that GPT-4.5\u2019s increased size has given it \u201ca deeper world knowledge\u201d and \u201chigher emotional intelligence.\u201d However, there are signs that the gains from scaling up data and computing are beginning to level off. On several AI benchmarks, GPT-4.5 falls short of newer AI \u201creasoning\u201d models from Chinese AI company DeepSeek, Anthropic, and OpenAI itself.<\/p>\n GPT-4.5 is also very expensive to run, OpenAI admits \u2014 so expensive that the company says it\u2019s evaluating whether to continue serving GPT-4.5 in its API in the long term. To access GPT-4.5\u2019s API, OpenAI is charging developers $75 for every million input tokens (roughly 750,000 words) and $150 for every million output tokens. Compare that to GPT-4o, which costs just $2.50 per million input tokens and $10 per million output tokens.<\/p>\n \u201cWe\u2019re sharing GPT\u20104.5 as a research preview to better understand its strengths and limitations,\u201d said OpenAI in a blog post shared with TechCrunch. \u201cWe\u2019re still exploring what it\u2019s capable of and are eager to see how people use it in ways we might not have expected.\u201d<\/p>\n OpenAI emphasizes that GPT-4.5 is not meant to be a drop-in replacement for GPT-4o<\/a>, the company\u2019s workhorse model that powers most of its API and ChatGPT. While GPT-4.5 supports features like file and image uploads and ChatGPT\u2019s canvas tool<\/a>, it currently lacks capabilities like support for ChatGPT\u2019s realistic two-way voice mode<\/a>.<\/p>\n In the plus column, GPT-4.5 is more performant than GPT-4o \u2014 and many other models besides. <\/p>\n On OpenAI\u2019s SimpleQA benchmark, which tests AI models on straightforward, factual questions, GPT-4.5 outperforms GPT-4o and OpenAI\u2019s reasoning models, o1<\/a> and o3-mini<\/a>, in terms of accuracy. According to OpenAI, GPT-4.5 hallucinates less frequently than most models, which in theory means it should be less likely to make stuff up<\/a>. <\/p>\n OpenAI did not list one of its top-performing AI reasoning models, deep research, on SimpleQA. An OpenAI spokesperson tells TechCrunch it has not publicly reported deep research\u2019s performance on this benchmark and claimed it\u2019s not a relevant comparison. Notably, AI startup Perplexity\u2019s Deep Research model, which performs similarly on other benchmarks to OpenAI\u2019s deep research, outperforms GPT-4.5 on this test of factual accuracy<\/a>.<\/p>\n On a subset of coding problems, the SWE-Bench Verified benchmark, GPT-4.5 roughly matches the performance of GPT-4o and o3-mini but falls short of OpenAI\u2019s deep research<\/a> and Anthropic\u2019s Claude 3.7 Sonnet<\/a>. On another coding test, OpenAI\u2019s SWE-Lancer benchmark, which measures an AI model\u2019s ability to develop full software features, GPT-4.5 outperforms GPT-4o and o3-mini, but falls short of deep research.<\/p>\n GPT-4.5 doesn\u2019t quite reach the performance of leading AI reasoning models such as o3-mini, DeepSeek\u2019s R1<\/a>, and Claude 3.7 Sonnet <\/a>(technically a hybrid model) on difficult academic benchmarks such as AIME and GPQA. But GPT-4.5 matches or bests leading non-reasoning models on those same tests, suggesting that the model performs well on math- and science-related problems.<\/p>\n OpenAI also claims that GPT-4.5 is qualitatively<\/em> superior to other models in areas that benchmarks don\u2019t capture well, like the ability to understand human intent. GPT-4.5 responds in a warmer and more natural tone, OpenAI says, and performs well on creative tasks such as writing and design.<\/p>\n In one informal test, OpenAI prompted GPT-4.5 and two other models, GPT-4o and o3-mini, to create a unicorn in SVG, a format for displaying graphics based on mathematical formulas and code. GPT-4.5 was the only AI model to create anything resembling a unicorn.<\/p>\n In another test, OpenAI asked GPT-4.5 and the other two models to respond to the prompt, \u201cI\u2019m going through a tough time after failing a test.\u201d GPT-4o and o3-mini gave helpful information, but GPT-4.5\u2019s response was the most socially appropriate. <\/p>\n \u201c[W]e look forward to gaining a more complete picture of GPT-4.5\u2019s capabilities through this release,\u201d OpenAI wrote in the blog post, \u201cbecause we recognize academic benchmarks don\u2019t always reflect real-world usefulness.\u201d<\/p>\n OpenAI claims that GPT\u20104.5 is \u201cat the frontier of what is possible in unsupervised learning.\u201d That may be true, but the model\u2019s limitations also appear to confirm speculation from experts that pre-training \u201cscaling laws\u201d won\u2019t continue to hold.<\/p>\n OpenAI co-founder and former chief scientist Ilya Sutskever said in December<\/a> that \u201cwe\u2019ve achieved peak data\u201d and that \u201cpre-training as we know it will unquestionably end.\u201d His comments echoed concerns<\/a> that AI investors, founders, and researchers shared with TechCrunch for a feature in November<\/a>.<\/p>\n In response to the pre-training hurdles, the industry \u2014 including OpenAI \u2014 has embraced reasoning models, which take longer than non-reasoning models to perform tasks but tend to be more consistent. By increasing the amount of time and computing power that AI reasoning models use to \u201cthink\u201d through problems, AI labs are confident they can significantly improve models\u2019 capabilities. <\/p>\n OpenAI plans to eventually combine its GPT series of models with its \u201co\u201d reasoning series, beginning with GPT-5 later this year<\/a>. GPT-4.5, which reportedly<\/a> was incredibly expensive to train, delayed several times, and failed to meet internal expectations, may not take the AI benchmark crown on its own. But OpenAI likely sees it as a steppingstone toward something far more powerful.<\/p>\n<\/div>\nMixed performance<\/h2>\n
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/02\/Screenshot-2025-02-27-at-11.12.37AM.jpg?w=680\")
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/02\/Screenshot-2025-02-27-at-11.11.00AM.jpg?w=680\")
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/02\/Screenshot-2025-02-27-at-2.18.08AM.jpg?w=680\")
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/02\/Screenshot-2025-02-27-at-2.24.39AM.jpg?w=680\")
Scaling laws challenged<\/h2>\n