OpenAI to Unveil GPT-5 with Enhanced AI Capabilities

OpenAI is scheduled to launch its highly anticipated GPT-5 model during a live event on Friday at 1 AM Beijing time. The company hinted at the release through a post on X, where the word "livestream" was altered to "LIVE5TREAM," subtly indicating the upcoming unveiling of GPT-5. This announcement has sparked significant interest within the tech community, as the new model is expected to surpass its predecessors in terms of capabilities and performance.

The live event is set to begin at 10 AM Pacific Time on Thursday, which corresponds to 1 AM Beijing time on Friday. This timing suggests that OpenAI is aiming to maximize global viewership, given the widespread interest in the new model. The company has also confirmed the release by updating its official website with a GPT-5 icon, further fueling speculation about the model's imminent launch.

The anticipation surrounding GPT-5 is not unfounded. The previous iterations of the model, including GPT-3.5 and GPT-4, have already made significant strides in the field of artificial intelligence. GPT-3.5, which powered the initial release of ChatGPT in November 2022, garnered global attention for its advanced language processing capabilities. OpenAI subsequently upgraded to GPT-4 in March 2023, further enhancing the model's performance. Since then, the company has continued to innovate, releasing several new models such as GPT-4o, GPT-4.5, GPT-4.1, and GPT-4-mini.

The release of GPT-5 is expected to build on these advancements, offering even more sophisticated features and improved performance. The model is likely to incorporate the latest developments in AI technology, making it a significant leap forward from its predecessors. This could include enhanced natural language processing, better contextual understanding, and more accurate responses to user queries.

GPT-5 is expected to be available in three versions: standard, mini, and nano. The standard version will serve as the primary model for ChatGPT and API, integrating the multimodal capabilities of the GPT series and the reasoning abilities of the o series. The mini version is positioned as a high-value option, focusing on fast response times. The nano version is designed specifically for API use, targeting local reasoning and embedded device applications.

According to early leaks, Horizon Alpha and Horizon Beta, which were available for early access on the OpenRouter platform, are likely to correspond to the GPT-5 nano and GPT-5 mini versions, respectively. Horizon Alpha boasts a 256K token context capacity and a generation speed of 130-150 tokens per second, excelling in creative writing and emotional understanding tasks. Horizon Beta, on the other hand, performs exceptionally well in UI design and front-end development, with similarly fast response times.

However, there are doubts about the technological breakthroughs in GPT-5. According to previous reports, the "Orion" project, initially planned as part of the GPT-5 release, was downgraded to GPT-4.5 due to underperforming expectations. This project did not achieve the revolutionary improvements over GPT-4o that were anticipated. As a result, the focus of GPT-5's development has shifted from pursuing groundbreaking technological advancements to emphasizing practical value and commercial returns. The primary selling points for commercialization will be improvements in programming and mathematical capabilities.

GPT-5 is expected to introduce a new reinforcement learning technology called the "universal verifier," which can automatically check and verify model responses to enhance accuracy. However, OpenAI currently faces challenges such as data bottlenecks and technical difficulties, which may lead to performance degradation when converting powerful base models into smooth conversational models. The scarcity of pre-training data and the resulting limitations on model capabilities will pose long-term challenges.

Despite these technical hurdles, OpenAI's user growth and revenue increases indicate a strong market demand for AI products. The company announced that the number of active users of ChatGPT has reached 700 million, with a year-on-year increase of over four times. The number of daily messages has exceeded 3 billion, and the number of paying business users has surged from 3 million in June to 5 million, an increase of over 66%, indicating a rapid increase in enterprise-level market acceptance. The company's annualized revenue has reached 120 billion yuan, a significant increase from approximately 40 billion yuan in 2024.

Despite the rapid user growth, OpenAI still lags behind Google's AI products. Google AI Overviews has approximately 2 billion monthly users, and the AI chatbot Gemini App has over 450 million monthly active users. The release of GPT-5 will be a critical test of whether the company can maintain its leading position in the fiercely competitive market.