OpenAI unveiled its 'doctor-level artificial intelligence (AI)' GPT-5 on the 7th (local time), signaling an intensified competition among high-performance AI models. The AI industry is accelerating technology development and infrastructure investment, aiming for artificial general intelligence (AGI) comparable to human intelligence.
On this day, OpenAI announced that its latest AI model, GPT-5, surpasses the existing flagship model 'GPT-4o' in terms of accuracy, speed, and expense efficiency. GPT-5 has integrated an upgraded version of 'GPT-4o' and the inference model 'o3'. Starting today, GPT-5 will be rolled out to all users of ChatGPT, which numbers around 700 million weekly. Free users can use GPT-5 within limits, while paid subscribers can enjoy more computational power and features.
GPT-5 has dramatically increased its response speed, and the unique phenomenon of AI generating inaccuracies, known as 'hallucination,' has also decreased, according to the company. The 'vibe coding' ability, which automatically codes when directed in natural language like English and Korean, has become more refined, and writing performance, including the ability to construct natural sentences, has improved. The tendency to speak only what users want to hear, referred to as 'flattery,' has also diminished.
Sam Altman, CEO of OpenAI, said, 'If GPT-3 felt like chatting with a high school student and GPT-4 felt like conversing with a college student, GPT-5 feels like having a conversation with a PhD-level expert on any topic.'
He noted that 'GPT-5 is an important step towards AGI,' but it has not yet reached the level of achieving AGI. While performance has improved, it means that GPT-5 has not yet advanced to the point of replacing humans. CEO Altman also acknowledged that the key factor needed for AI to reach parity with humans is a lack of 'the ability to learn independently.'
Reactions to GPT-5 have been mixed. Reuters reported that while evaluations from pre-experience testers noted GPT-5 excels in coding and solving scientific and mathematical problems, the performance enhancement is not as groundbreaking as the leap from GPT-3 to GPT-4.
MIT Technology Review assessed, 'If the inference model 'o1' was a significant technological leap, GPT-5 is a refined product that maximizes user convenience.' During a press conference, CEO Altman compared GPT-5 to Apple's 'Retina Display,' suggesting that like an exceptionally clear screen, the latest AI model offers a more convenient and seamless user experience. MIT Technology Review pointed out, 'However, this is far from the 'transformative future led by AI' that CEO Altman has heavily promoted over the past year.'
Chirag Dekate, a researcher at market research firm Gartner, said, 'GPT-5 has exceeded expectations in inference and coding and has shown gradual progress in writing quality, but it has not crossed the threshold for AGI.' In Korea, the results from using ChatGPT based on GPT-5 for the 2025 College Scholastic Ability Test showed scores of 95 and 82 in Korean and English, respectively, surpassing the first-grade cut-off, while a score of 92 in mathematics fell into the second-grade range.
Competitors of OpenAI are also rushing to release high-performance AI models aimed at achieving human-level AGI. Anthropic, dubbed the 'OpenAI challenger,' introduced a cutting-edge AI model called 'Claude Apus 4.1' that significantly improved coding performance ahead of the GPT-5 release on the 5th. Google DeepMind also launched 'Gemini 3' on the 5th, capable of creating 3D virtual worlds in real-time when directed in natural language. Google stated that 'Gemini 3' marks a point where AI can move beyond simple video and text generation to learn through direct experience like humans, suggesting it will be a key turning point in the journey toward AGI.
Elon Musk, CEO of Tesla, is leading xAI, which is preparing for the launch of its cutting-edge AI model 'Groq 5' by the end of the year. After the release of GPT-5, Musk posted on X (formerly Twitter), 'xAI's 'Groq 4 Heavy' was smarter than the current GPT-5 two weeks ago, and it is certainly superior now,' indicating competitive pressure. The parent company of Facebook, Meta, announced plans to establish a dedicated research center and undertake large-scale investments aiming for 'superintelligent AI' beyond AGI.
There are also analyses suggesting that as the performance of AI models generally rises, it may become difficult for OpenAI to maintain its leadership position in the AI market. AI models follow the principle of 'scaling,' using vast computational resources and high-quality data for performance enhancement, which incurs enormous expenses when training and operating even a single latest AI model. Startups like OpenAI, which need to secure funding from external sources, are inevitably at a relative disadvantage in competition compared to big tech companies with significant financial resources. CEO Altman also emphasized that 'constant scaling' is necessary to maintain a competitive edge in the AI market.
He stated, 'The next leap will require an enormous level of computing resources, but we will continue to push forward.'